How diagnostic tests work in the field — and why only a single date is not gonna work
Over the course of the Covid Mapping Project, I’ve been warning people not to read too much into new case counts — whether the trends look positive or negative. I’ve been urging caution about this because we simply do not know enough about what testing is being done, where, by whom, and most importantly, when those tests are done. I know a little bit about this because my day job at Standard Co is helping run a data platform (Secure Data Kit) that processes diagnostic test results for tens of thousands of field surveys in over 60 countries.
About the data Georgia is reporting
You can see that twice each day at noon and 7pm ET, Georgia reports how many tests cumulatively have been run.
That’s all they tell us...
Based on this information, we have no idea if the patient was tested 12 hours ago or 12 days ago. It is entirely possible that the patient had Covid and is well on their path to recovery by the time the data is reported.
The data I would like to see
Based on how we design our diagnostic data capture systems, there are a few simple things I would like to see the state of Georgia report on:
- When was the sample taken: this is *critical*. It helps us understand the point in time when the sample was actually gathered
- When was the sample transmitted to the lab: not quite as critical as the above. But it helps us understand if there is any latency between the ones conducting the test and the actual analysis of the sample
- When was the sample analyzed: depending on the study, where the lab is, and what country all of this is happening in, our diagnostic results can take a day or a week or even longer. Again, these are studies for diseases not as viral as COVID-19 nor quite as deadly. So the time factor isn't as important (but still important)
Other data that would be extremely helpful to know:
- How many samples transmitted on x date
- How many samples awaiting testing
- Testing capacity per business day / week day
These kind of data points would help us get a sense for throughput of the lab. And it would help prepare us if an onslaught of positive tests were to head our way.
I would be shocked if the state didn’t have this kind of data hiding under the hood. For one reason or another, they’ve decided to provide a single date — the date they publish the data. Without the other date points, we’re merely guessing as to the timeliness of the data we’re looking at.