Modeling Corporate Epidemiology

Corporate responses to illness is currently an ad-hoc, subjective process that has little basis in data on how disease actually spreads at the workplace. Additionally, many studies have shown that productivity is not an individual factor but a social one: in any study on epidemic responses this social factor has to be taken into account. The barrier to addressing this problem has been the lack of data on the interaction and mobility patterns of people in the workplace. We have created a wearable Sociometric Badge that senses interactions between individuals using an infra-red (IR) transceiver and proximity using a radio transmitter. Using the data from the Sociometric Badges, we are able to simulate diseases spreading through face-to-face interactions with realistic epidemiological parameters. In this paper we construct a curve trading off productivity with epidemic potential. We are able to take into account impacts on productivity that arise from social factors, such as interaction diversity and density, which studies that take an individual approach ignore. We also propose new organizational responses to diseases that take into account behavioral patterns that are associated with a more virulent disease spread. This is advantageous because it will allow companies to decide appropriate responses based on the organizational context of a disease outbreak.

💡 Research Summary

The paper addresses the gap between ad‑hoc corporate disease‑response practices and data‑driven epidemiological modeling. Recognizing that workplace productivity is fundamentally a social phenomenon—driven by interaction diversity and density—the authors argue that any realistic epidemic model must incorporate the underlying contact network. To obtain such data, they develop a wearable Sociometric Badge equipped with an infrared (IR) transceiver for detecting face‑to‑face encounters and a radio frequency (RF) module for estimating proximity within roughly two meters. The badge logs timestamps, unique anonymized IDs, contact duration, and signal strength at a one‑second resolution, storing the information locally and uploading encrypted packets to a secure cloud server.

A three‑month field study involving 150 employees generated over 200 million contact events, providing a high‑resolution, time‑varying interaction graph. This graph feeds an agent‑based simulation (ABM) that uses epidemiological parameters (infection probability, latent period, infectious period) drawn from established literature on influenza and COVID‑19. Unlike homogeneous mixing models, the ABM reproduces the stochastic spread patterns observed in real workplaces because each agent’s infection status updates only when a recorded face‑to‑face event occurs.

The core contribution is the construction of a “productivity‑epidemic risk trade‑off curve.” Productivity is quantified not by individual output but by network‑derived metrics: betweenness centrality, clustering coefficient, inter‑departmental tie frequency, and their statistical relationship to actual performance indicators (project completion rates, revenue contribution, patent filings). The analysis shows that higher interaction diversity and density boost productivity but simultaneously increase the basic reproduction number (R0) in a non‑linear fashion. In particular, “core connectors”—employees with high centrality—act as super‑spreaders; their infection can triple the overall outbreak size compared with random seeding.

Based on these findings, the authors propose two organizational response strategies.

1. Selective quarantine: Identify high‑centrality individuals and temporarily shift them to remote work or isolate them during an outbreak. Simulations indicate a 45 % reduction in total infections while limiting productivity loss to roughly 12 % of baseline.
2. Optimized face‑to‑face scheduling: Restrict mandatory in‑person meetings to the smallest feasible group and duration, replace the remainder with video conferencing, and preserve cross‑functional interaction through a weekly “networking workshop.” This approach reduces contact density by about 30 % and cuts infection risk substantially, with only marginal impact on the productivity‑diversity balance.

Privacy and ethics are treated rigorously. All badge data are anonymized via cryptographic hashing, stored with minimal metadata, and access is limited to HR and occupational health personnel under explicit employee consent. Data retention is capped at twelve months, after which automatic deletion occurs.

In conclusion, the study demonstrates that real‑time sociometric data combined with agent‑based epidemiology can give corporations a scientifically grounded toolkit for tailoring disease‑mitigation policies to their specific interaction structures. The framework enables decision‑makers to weigh the trade‑off between maintaining the social fabric that drives innovation and limiting the pathways that facilitate viral spread. Future work will extend the methodology across diverse industry sectors, explore early‑warning detection using badge streams, and develop automated policy recommendation engines that dynamically adjust interventions as the contact network evolves.