Walk Before You Run: Starting with Reliable and Clean Descriptive Data to Address COVID-19

Walk Before You Run: Starting with Reliable and Clean Descriptive Data to Address COVID-19

Jul 28, 2020
CoviSecure sign in window

It is March 2020. It’s becoming clear as the days progress, that COVID-19 will not go away. In fact, until a vaccine is found, it’s here to stay and workplaces, especially hospitals and other providers who deliver essential medical care, need to figure out their response. For these institutions, shutting down was not an option.

Enter our colleagues at OnPacePlus. They reached out to Mathematica early in March to discuss innovative tools they were working on to respond to COVID. Specifically, they were interested in predicting the likelihood of COVID infection from reported symptoms. As we collaborated, it became clear to us in those early days that the emerging science on COVID was not mature. The data were unavailable, and reliable prediction via symptomatology would be very difficult. Mathematica is well known for our ability to leverage advanced analytics to deliver value to our clients but as with any endeavor, it is vital to meet the institution at its current state—overengineering a solution can be just as detrimental as providing no solution at all.

We took a step back and asked ourselves, What if we don’t lead with prediction? What if we walk before we run? Using the relationships we’ve built over the years, we contacted our physician colleagues and focused on the local areas where COVID hit the hardest. We spoke to providers in hot spots like Westchester, New York, and across Philadelphia’s Main Line. It became clear early on that the market was not ready for prediction or advanced analytics. Instead, health care providers needed a mechanism—a descriptive statistical tool—to report the symptoms their staff were experiencing, their staff’s access to personal protective equipment, and access to ventilators and their critical components. Further, collecting these data needed to be easy enough that it did not add burden to providers’ already hectic workflows. The data needed to be cleaned and visualized in an intuitive way that was useful for decision making. Finally, the data needed to get in front of decision makers (for example, the chief medical officer or director of occupational medicine) so that evidence rather than anecdote could be leveraged for workplace management. Walk before you run became our theme.

Mathematica and OnPacePlus knew we could meet this challenge, so we partnered on developing a system to meet the market at its current state. We developed a COVID-19 symptomology survey leveraging the clinical expertise in medicine and infectious disease at OnPacePlus and the survey methods expertise for questionnaire design from Mathematica. The survey is fully mobile, easy to use, and can be completed in fewer than 10 minutes. With the survey operational, we then focused on our extract, transform, and load processes for moving the survey data into a fully secure data warehouse where the data can be cleaned and organized. Finally, we worked through the development life cycle to stand up a visual analytics platform that leveraged the principles of user interface design to ensure data interpretation by administrators was seamless. After the platform was up and running, administrators were able to use their data to ask critical workforce management questions: (1) who is available to work, (2) what symptoms are my staff experiencing, and (3) how are COVID symptoms spread across locations and employee type? Achieving well-organized and clean data was the first step to answering these questions and was consistent with our mindset of walk before you run.

As the team looks ahead, we are considering more advanced modeling strategies. For example, we are working to customize Mathematica’s 19 and Me calculator to assess COVID-19 risk for physician providers. This customization will enable stakeholders to understand the differential risk between an emergency room physician and an ophthalmologist with respect to COVID-19. Moving on to this next phase of advanced modeling would not have been possible without our early focus on establishing clean and reliable descriptive statistics. Now that the team has walked, we’re ready to run.

About the Author

Prevent, detect, and respond.

Address public health issues with the data, evidence, and the policy expertise you need to improve public well-being.

Learn More