Household–Workplace–School (HWS)

Life in Layers: How the Household–Workplace–School Model Captures the Rhythm of Everyday Epidemics

Simulating disease spread through the daily dance of home, work, and school

🏠 Introduction

Your day has a rhythm: wake up with your family, head to work or school, grab coffee with a colleague, maybe swing by the gym. Each stop is a microcosm of contact—intimate at home, structured at work, chaotic in the community. Now imagine a virus slipping into this routine. It doesn’t spread randomly; it rides the rails of your daily life.

This is the insight behind the Household–Workplace–School (HWS) agent-based model—a framework that structures epidemic simulations around the three pillars of human interaction. Unlike models that treat society as a uniform soup or a static web, HWS recognizes that who you infect depends on where you are.

Developed during the 2000s to model influenza and refined for SARS-CoV-2 [1–3], HWS models have become a gold standard for evaluating real-world interventions: school closures, remote work, household quarantine. They’re detailed enough to capture superspreading in classrooms, yet efficient enough to run thousands of scenarios.

In this article, we’ll unpack how HWS models work, why their layered design mirrors reality, and how scientists extend them to tackle everything from measles outbreaks to pandemic preparedness.

🔄 Model Description

The HWS model assigns each agent a daily schedule that cycles through three (or four) key settings:

  1. Household (H): Night and early morning
  2. Workplace or School (W/S): Daytime
  3. Community (C): Evenings or weekends (optional)

Each agent belongs to exactly one household, one workplace (adults) or school class (children), and may participate in community mixing.

Agents progress through a SEIR disease state:

  • S: Susceptible
  • E: Exposed
  • I: Infectious
  • R: Recovered

Time advances in daily steps. On each day, transmission is evaluated separately within each setting the agent visits.

🦠 Transmission Rule

For each setting σ ∈ {H, W/S, C}, a susceptible agent j interacts with a set of contacts Cⱼ⁽ᵟ⁾ drawn from their assigned group in that setting. For every infectious contact i ∈ Cⱼ⁽ᵟ⁾ ∩ I, infection occurs with probability:

pᵢⱼ⁽ᵟ⁾(t) = 1 − exp(−βᵟ · g(τᵢ))

Where:

  • βᵟ is the setting-specific transmissibility scale (e.g., βₕ > β꜀ due to closer contact),
  • g(τ) is the infectiousness curve (peaking early in infection),
  • τᵢ is time since agent i became infectious.

The total daily infection probability for agent j is the complement of escaping infection in all settings:

Pⱼ(t) = 1 − ∏ᵟ ∏ᵢ∈Cⱼ⁽ᵟ⁾∩I (1 − pᵢⱼ⁽ᵟ⁾(t))

This captures the idea that infection can occur at home, at work, or in the community—and risk compounds across settings.

👨‍👩‍👧‍👦 Contact Group Assignment

  • Households: Agents are grouped into units of size h, drawn from empirical distributions (e.g., mean = 2.6 in the U.S.).
  • Schools: Children assigned to classes of size c (e.g., 20–30 students).
  • Workplaces: Adults assigned to teams or departments of size w (e.g., 10–50).
  • Community: Optional random mixing with a subset of the population (e.g., 5–15 random contacts per day).

Contact groups are fixed during the simulation (unless interventions change them).

⏳ Disease Natural History

Each agent draws:

  • Tᴱᵢ ~ incubation period (e.g., log-normal, mean 2–5 days),
  • Tᴵᵢ ~ infectious period (mean 3–10 days).

Progression: S → E → I → R.

📊 Key Parameter Definitions & Typical Values

NPopulation sizeTotal agents10,000 – 10⁷
⟨h⟩Mean household sizeAvg. people per home2 – 4
⟨c⟩Mean class sizeStudents per classroom15 – 30
⟨w⟩Mean workplace sizeColleagues per team10 – 50
βₕHome transmissibilityPer-contact risk at home0.1 – 0.5
βᵥWork/school transmissibilityRisk in structured settings0.05 – 0.3
β꜀Community transmissibilityRisk in casual settings0.01 – 0.1
g(τ)Infectiousness curvePeaks at τ = 1–3 daysBell-shaped
⟨Tᴱ⟩Mean incubationTime to infectiousness1–10 days
⟨Tᴵ⟩Mean infectious periodDuration of infectiousness3–14 days

💡 Why different β? Close, prolonged contact at home (shared air, surfaces) makes βₕ higher than βᵥ or β꜀ [4].

⚖️ Assumptions and Applicability

The HWS model makes several realistic simplifications:

✅ Structured daily routine: Agents follow fixed schedules (valid for stable populations).
✅ Homogeneous mixing within groups: Everyone in a class or household has equal contact chance.
✅ No cross-group workplace mixing: Colleagues only interact within their team (relaxed in variants).
✅ Age stratification: Children go to school; adults to work—enabling age-targeted policies.

🎯 When is this model most useful?

  • Respiratory diseases with strong setting dependence: Influenza, RSV, SARS-CoV-2, measles [1,5].
  • Evaluating non-pharmaceutical interventions (NPIs):
     – School closures
     – Remote work mandates

 – Household quarantine

  • Urban or national-scale planning: When demographic and contact data are available [6].
  • Age-specific risk analysis: Children vs. adults in transmission dynamics [7].

It is less suitable for:

  • Diseases with environmental transmission (e.g., cholera),
  • Highly mobile or homeless populations,
  • Settings where social networks dominate over institutional ties (e.g., close-knit villages).

🔧 Model Extensions and Variants

To enhance realism and policy relevance, researchers have developed key extensions:

1. Layered Contact Intensities

Instead of binary contact groups, assign contact durations or frequencies:

pᵢⱼ⁽ᵟ⁾(t) = 1 − exp(−βᵟ · dᵢⱼ⁽ᵟ⁾ · g(τᵢ))

where dᵢⱼ⁽ᵟ⁾ is daily contact time (e.g., 8 hrs at home, 1 hr at work) [8].

2. Asymptomatic & Presymptomatic Transmission

Introduce fraction fₐ of asymptomatic cases with modified infectiousness:

gₐ(τ) = α · g(τ), α < 1

Asymptomatic agents may have different βᵟ (e.g., lower if they stay home) [2].

3. Dynamic Group Disruption

Model interventions by altering group structure:

  • School closure: Set Cⱼ⁽ˢ⁾ = ∅ for children,
  • Remote work: Move adults from workplace to household-only mixing,
  • Household quarantine: Remove infectious agents from all non-household contacts [9].

4. Age-Structured Susceptibility & Severity

Assign age-dependent parameters:

  • sₐ: Susceptibility (e.g., higher in children for RSV),
  • hₐ: Hospitalization risk (e.g., higher in elderly).

Enables realistic burden forecasting [7].

5. Community Mobility Networks

Replace random community mixing with activity-driven or spatial models for the community layer, capturing local hotspots [10].

These variants make HWS models policy-ready tools used by CDC, WHO, and national health agencies.

🎉 Conclusion

The Household–Workplace–School model succeeds by mirroring a simple truth: epidemics unfold in the spaces of everyday life. By layering transmission across home, work, and school, it captures why closing schools slows flu—or why working from home reduced SARS-CoV-2 spread.

Its power lies in balance: detailed enough to reflect social structure, yet modular enough to test interventions with surgical precision. During the 2009 H1N1 pandemic and the 2020–2023 COVID-19 crisis, HWS models provided the backbone for real-time decision-making in dozens of countries [1,5,9].

In an era of complex outbreaks, the HWS model reminds us that sometimes, the best way to stop a pandemic is to understand the rhythm of ordinary days. Because in the end, it’s not just about viruses—it’s about where we live, where we work, and who we sit next to in class.

So the next time you hug your child, share a coffee with a coworker, or ride the bus home, remember: you’re not just living your life. You’re moving through a layered network of risk—and in the eyes of an epidemiologist, every layer tells a story.

📚 References

  1. Ferguson, N.M. et al. (2005). Strategies for containing an emerging influenza pandemic in Southeast Asia. Nature. https://doi.org/10.1038/nature04017
  2. He, X. et al. (2020). Temporal dynamics in viral shedding and transmissibility of COVID-19. Nature Medicine. https://doi.org/10.1038/s41591-020-0869-5
  3. Mossong, J. et al. (2008). Social contacts and mixing patterns relevant to the spread of infectious diseases. PLOS Medicine. https://doi.org/10.1371/journal.pmed.0050074
  4. Glass, K. et al. (2022). Agent-based models for infectious disease policy. Philosophical Transactions B. https://doi.org/10.1098/rstb.2021.0030
  5. Cauchemez, S. et al. (2009). Closure of schools during an influenza pandemic. The Lancet Infectious Diseases.
  6. Prem, K. et al. (2017). Projecting social contact matrices in 152 countries. PLOS Computational Biology. https://doi.org/10.1371/journal.pcbi.1005697
  7. Davies, N.G. et al. (2020). Age-dependent effects in the transmission and control of COVID-19. Nature Medicine. https://doi.org/10.1038/s41591-020-0962-9
  8. Jarvis, C.I. et al. (2020). Quantifying the impact of physical distancing on contact patterns. The Lancet Infectious Diseases.
  9. Kerr, C.C. et al. (2021). Controlling COVID-19 via test, trace, and isolate. Nature Communications.
  10. Aleta, A. et al. (2020). Modelling the impact of testing and contact tracing on second waves. Nature Human Behaviour. https://doi.org/10.1038/s41562-020-0931-9