Why don’t we have a National Dashboard?
Is the coronavirus spreading faster than before? Are there more very sick people now than sometime back? Is the virus as dangerous now as it was at the start? Which places are getting worse, by how much, and, more importantly, why? Did the lockdown really work, and if so how well did it work in each place?
Six months into a pandemic we should have precise answers to all these questions. Particularly because the answers drive major decisions that affect countless lives and trillions of dollars of economic impact. The good news is that the answers to most of these questions depend on data — supposedly an objective thing. Yet, six months (we think) into the pandemic, we still aren’t sure if the data we have across states is semantically consistent and clinically meaningful.
Two examples to think about. Test positivity rate may have gone up in (state) A and down in B. Does this mean the virus is spreading faster in A? What if A was pursuing the most sick cases for testing that week while B was using random testing open to anyone? Should that change anything? Of course it should.
Problem is, week by week, we have no idea of the true process behind these numbers and play this guessing game. And, in case you didn’t notice, Florida recently changed the way ICU capability is reported to its tracking system.
Detractors claimed this was an effort to under-report. Proponents claimed that some hospitals were using regular ICU beds as a “COVID wing” and that they wanted to capture the actually serious cases. Regardless, dashboards will treat this as the same number semantically across weeks. What would this do to week-by-week comparisons?
But don’t worry, Johns Hopkins figured it all out and built the perfect COVID-19 dashboard. Right? A recent Washington Post article had this to say: “Case counts are consistently inconsistent. Reporting practices differ from country to country, state to state, even county to county.” These statements weren’t meant as criticisms of the Hopkins team. They actually reported the worries and sentiments of those brilliant minds there that put this together.
Given how important data on this virus is, why don’t we have a national dashboard, with a set of clearly defined metrics, identified by doctors and health experts, that will be reported on by every clinic and hospital in the country directly to a central national database? This would include a number based on a subset of random individuals so we have proper baselines of the presence of the virus in the population instead of guessing based on how the wind blows each week on whom states decide to prioritize to test.
Experts would have defined metrics that could capture the intensity of the sickness, the spread and processes that need to be in place. After that it would have been a question of execution, to require counties and states to comply with these reporting guidelines, along with audits to ensure the correct data was being supplied. Sure, we would have had to support the hospitals and clinics in some manner, personnel wise, to provide this service. There are about 7,000 hospitals in the country. If we had funded three full time hires in each of these places dedicated to these tasks it would have cost us about $1.5 billion. Expensive, yes. But given the cost of this pandemic this is less than a rounding error.
We have for years known the importance of standardization and getting the semantics (or meaning) of data correct and consistent. It’s a shame that six months and trillions of dollars
into a pandemic we are still interpreting data any way that suits our perspective. A unified, single version of truth, would have consistently told one story — the right one — so that all our decisions along the way would have been calibrated precisely to what the data would have called for.
Are the current dashboards useful? Absolutely. They are good at highlighting strong trends. They clearly bring out the fact that we may be done with the virus, but the virus isn’t done with us. Thirty-one states in the United States reported higher new cases this week compared to last week. The virus has clearly affected some socioeconomic groups more than others. When we are (ever) done with the virus, then tests will still result in a low number of cases, leading to negligible test positivity rates. The dashboards, even as constructed, will help us see this.
But can we do better? Of course we can, and of course we should. I will call for a higher bar for data consistency and data quality, and for a nationally coordinated dashboard with consistent semantics and clinically powerful metrics. Not to move the needle again in the middle of the pandemic, but to get better prepared when the next one hits. Because we know it will.
Balaji Padmanabhan is a University of South Florida Muma College of Business professor who teaches courses in areas related to artificial intelligence and machine learning, business/data analytics and computational thinking. He is also the director of the Center for Analytics & Creativity.