As we enter the autumn, workers around the world are returning to the office. Deaths from COVID-19 cases continue to fall from their spring peaks in Europe and North America, and governments are permitting greater levels of social interaction. Organizations are also beginning to observe the value of the collaboration, mentorship, and culture creation that proximity generates.
One thing is clear as property managers and enterprise businesses try to navigate the re-entry of employees and customers: The guidelines issued by the CDC and other governmental organizations are too broad and do not help people understand their personal risk within a specific environment.
At Haven Diagnostics, we are working with commercial property owners using a relatively new science, agent-based modeling. This new technique has been made possible by increased computing power and is transforming the way epidemiologists approach infectious disease outbreaks. Agent-based models allow individual rule-based agents, each with their own variables, to interact with each other dynamically. Epidemiologists can thereby simulate real-world complexity and randomness in a sophisticated way, helping with risk management. The National Institutes of Health has labeled agent-based models a “disruptive innovation” and, within the past year, has funded scientists at Harvard, Columbia, and Yale to develop agent-based models for use against hospital-acquired infections, for defense against bioterrorism, and for combating the spread of influenza in the workplace.
The problem is inherent in the fact that each person, and by extension, each property, has its own risk profile, and any mitigation efforts need to take those into consideration. Simply put, the actions that a nursing home in Manhattan needs to take to re-allow visitors are far different than the ones needed on a college campus in Illinois.
Thankfully, we can now address this problem. Agent-based computational models simulate the interactions of independent agents (people in a hotel lobby, for example) with each other, and across different environments (with higher or lower density restrictions). This allows asset and property managers, as well as business operators to develop a deeper understanding of their idiosyncratic risk profile, and which actions are the cheapest and most effective way to reduce this risk.
The animation below shows a group of people interacting with each other and sharing a common bathroom. Each red dot is a person (agent) infected with COVID-19 who can pass the disease to others, and the blue dots are susceptible people. Each person has different characteristics that determine how likely they are to contract COVID-19, and to pass it on to their peers (their level of interaction with others, whether they wear masks, how long they stay in the facility, etc.).
This approach can be scaled and applied to different layouts, demographics, and populations, thereby simulating the impact of different scenarios. Within each property, the first step is to integrate different variables to give each agent its appropriate characteristics. We integrate four categories of variables: demographics, behavior, building systems and layout, and location information. Our statistical programming team takes into account a wide array of big data including CDC data at a sub-zip-code level, HVAC system specifications, and over 500 medical studies (such as impact of droplet spread in bathrooms) in order to determine the likelihood of disease transmission in a property.
The schematic below is an example of how we can simulate disease spread in an office environment, and recommend a floor plan and density schedule that allows an executive to manage re-entry and stay open through a variety of scenarios (a second wave, for example).
With my background in applied artificial intelligence, I am keenly aware that a “one size fits all” approach is insufficient to manage the complexity of variables that affect workplace re-entry during a pandemic. A thoughtful statistical approach, incorporating demographic, layout, and facility data that is specific to each site is critical to assessing risk and truly reassuring employees and tenants. Thanks to recent advances in statistics and the ready availability of workplace data, we do have the platform to reopen our workplaces, reassure all stakeholders, and most importantly, avoid the outbreaks that could cause us to shut down again in the autumn and winter.