NY Edition: Daily COVID-19 Data Is About to Get Weird
The COVID Tracking Project’s blog featured an interesting post yesterday, trying to understand how the Thanksgiving holiday might impact COVID numbers — but not in the way you might think. They weren’t analyzing whether another spike in cases would occur because of increased travel. Instead, they were understanding how COVID data reporting changes on holidays. The whole post is worth a read.
Earlier this week I wrote about some of the ways New York State has reported it’s COVID data versus the way Onondaga County has reported. The COVID Tracking Project’s post got me wondering how New York State and Onondaga County’s data compares.
The key takeaway from the COVID Tracking Project’s blog post, and what I’ll try to look at here, is if there is the potential for an dip in reported COVID cases over the next few days that make it look like the worry about Thanksgiving gatherings was overblown, followed by an very steep climb in cases, causing people to blame Thanksgiving. In actuality, it might just be about when cases are reported.
State Level
When looking at data for all of the State of New York in the visualizations above, you can see that the individual bars of the bar chart are above and below the 7-day rolling average of cases and tests with a pretty consistent pattern. This makes sense since the average broadly at the week’s numbers, while individual days can fluctuate. The pattern, as we’ll see, shows that fewer cases are reported on certain days of the week very consistently, while more cases are reported on other days with consistency as well.
In the visualization above, looking at both case (in red) and test (in blue) reporting for the state overall, the bars that show a percentage greater than 0 mean that the 7-day rolling average is greater than that days reports. For bars that are less than zero, it means the daily reports were more than the 7-day average. We can see that Sundays and Mondays report much fewer cases and tests than the 7-day average, and on Wednesday, Thursday, and Friday, the daily cases and tests are greater than the rolling average. Since the Thanksgiving holiday falls on Thursday, many of the testing sites may not be open (this could be true for Friday, too, depending on the location), meaning the days then the tests and cases catch up to the rolling average likely won’t catch up this week. Initially, the data may make it seem like the cases have suddenly dropped, which would seem like good news. By next week, though, people will be getting tested again and labs will report test results, including those that would have been reported on Thursday and Friday if not for the holiday. So, we might see a large drop in cases followed by a large increase in tests and cases, neither of which reflect reality of disease spread, but instead reflect which days the labs were open.
By County
I wanted to see if these trends are consistent by count within New York State. Because many counties across the state have relatively few cases, I only looked at the top 15 counties. The trend of case and test reporting per day being less than the 7-day rolling averages on Sundays and Mondays with Wednesdays-Fridays serving as the days where the daily cases are more than the average. It is interesting to see that the reporting differences are not consistent across counties each day. In Albany County, Sunday case reporting is much less than the 7-day average (more than 40% difference), while in Queens County the difference is only about 20% on Sundays. This could be based on a number of factors — how many cases and tests are administered overall in the county as well as the reporting frequencies.
State Reports for Onondaga County
Specific to Onondaga County, according to New York State reported data, the weekend underreporting of daily cases versus the 7-day rolling average remains true, although it is interesting that reports of daily tests are greater than the 7-day average on Mondays, while reporting of positive cases daily lags behind the 7-day average.
Because Thursdays and Fridays are generally when daily test and case reporting exceeds the 7-day average, again in Onondaga County it could initially look like case surge is lessening and then be followed by what looks like a surge in cases. We should not assume that potential surge is associated with Thanksgiving gatherings. Gatherings might well lead to additional case spread, and we should absolutely be adhering to CDC guidelines and staying home, but the data might not align with reality initially.
County Numbers
Up until this point, we’ve looked at data reported by the state, now we will look at data reported by Onondaga County. Interestingly, the days when cases are reported is very different from the State’s reporting. Monday and Tuesdays are the days when County cases lag behind the 7-day average. Similar to the State data reporting, though, Thursday and Fridays are typically days where daily case numbers are greater than the 7-day rolling average. So, we might expect a similar trend where cases suddenly drop and then increase. Especially given that the County just reported its worst single day of cases yesterday at 270, attention to continued spread without getting distracted by reporting weirdness over the next few days.
County Numbers Hospitalizations
Finally, hospitalization reports should see daily hospitalizations closer to the 7-day rolling average because hospitals are open all the time and if someone needs to go to the hospital, they won’t wait until a specific day. Above, the green bars show that this is true when looking at the total number of people hospitalized. For new hospitalizations, the differences are much greater. I think this is because Onondaga County has generally had a low number of hospitalizations (thankfully), so any days with a jump of a few additional cases would have an impact on the overall numbers.
From this perspective looking at how total hospitalizations changes over the next few weeks will likely be the best indicator to how Thanksgiving gatherings impact case spread in Onondaga County. Though, since hospitalizations typically lag behind positive tests by a couple of weeks, we’ll need to be patient in analyzing that data.
Overall, all of this is to say that using data to analyze COVID-19 trends is important to understand the policies to implement to help control spread. Understanding this though, this data is not ground truth. The experts working on this every day who understand the context for the data can help to explain what is happening.
