Ben Kemp, Principal Scientist, explains how LabGenius has put data at the heart of decision-making to help navigate the winding road back to the office.

The COVID risk model discussed in this post has been made accessible to all via an easy-to-use online tool so that anyone can update the input fields and make the output relevant to different hybrid working environments. We recognize that the model is in its infancy but our aim is to promote collaborative problem solving amongst those that are facing similar challenges. As we continue to evolve the tool, we welcome any feedback, so please respond to the blog post or contact us directly via covidmodel@labgeni.us

As the traditional office setting continues to find its place in a post-COVID world, we want to embrace the flexibility that hybrid working brings while ensuring that we maintain business continuity and keep the wider team safe.

To achieve this, we have landed on a flexible post-COVID working strategy that has one central expectation: everyone should come into the office at least twice a week to allow for those all-important team interactions that we’ve missed so much.

What did LabGenius’ working environment look like during the pandemic?

As a company that operates at the intersection of technology and science, our wet-lab team has been coming into the office throughout the pandemic — after all, you can’t engineer proteins in your kitchen. To keep the wet-lab team safe, the rest of the team has worked from home wherever possible to minimize the number of people on site.

So, what do we need to consider when planning a safe return to the office?

The biggest challenge we face is deciding when to launch our post-COVID working strategy. The UK government lifted most restrictions on the 19th of July, but to keep our team safe and ensure we keep our wet-lab running we decided to open up more carefully. These points were front of mind:

• With 64% of employees between the ages of 25–34, we’ve got a relatively young team, and in the UK the rollout of the vaccine is slower in younger populations. However, having a younger team also means that we are less likely to see serious cases of COVID.
• We’re based in London, which has a high population density, increasing the chance of meeting someone with COVID on your way to work.
• As a small company, even one or two people self-isolating in our wet-lab team has the potential to impact how much we can achieve.

What does our practical, data-driven model for returning to the office look like?

As a deep tech company, it was only right that we take a proactive, data-driven approach to our return to work strategy. Using some readily available information, we have developed a COVID risk model that has helped us to identify the point at which we are comfortable asking the whole team to return to the office two days a week. The model has been made accessible to all via an online tool so that you can update the input fields and make the output relevant to any hybrid working environment. We used Causal, which is an awesome tool for simulation, model-building, and exploration.

Here are some of the finer details and assumptions that were used to develop our COVID risk model.

The key decision is: ‘how many infectious interactions are we willing to expose members of our team to per week?’ We’ve seen the effect of one or two positive cases on our relatively small team, and we don’t want to replicate those situations — so, we’ve made the decision to keep the number of infectious interactions below 1 per week.

We wanted a model, based on this number, that would help us figure out what the local case rate (per 100,000) should be for us to consider it safe to unlock. For us, “local” means London, where there are 280 cases per 100,000 for the week commencing 17th August.

Our model combines this acceptable number of infection interactions per week with the total number of days our team will be in the office and an estimate of the number of people they will interact with on their commute in a given week.

It also takes the efficacy of double-dose vaccination for preventing infection into consideration (65% or more, according to Public Health England) and the vaccination rate within our own team, which we recently and anonymously surveyed (75% and rising).

The model uses Monte-Carlo simulation to accurately account for uncertainty about ranges of possible values. Often when modeling with data, large ranges of possible outcomes are obscured by using average, but not representative, values. Using probabilistic simulation helps us understand the full range of possible outcomes for our team.

Based on the above, our model tells us that it’s safe to bring everyone back to the office for between 2 and 5 days each week, only once the case rate per 100,000 in London drops below 100. There are some uncertainties inherent in the calculation which means the output value, 100, falls within a wide range. Here we have chosen to go with 100 cases, which is approximately the 10th percentile value of this range, meaning that — cutting through all the uncertainty — there is a high chance the team will be safe if cases are below 100.

We’re not epidemiologists and this model is basic but it has helped us put personalized parameters in place and provided some certainty for our return to the office strategy. Our aim is to continually evolve the tool, so we would welcome responses via Medium. Alternatively, you can email us directly at covidmodel@labgeni.us — we would love to chat!