Our blog post about the major change in Florida’s COVID-19 reporting to the CDC has developed into a national story with partisan biases. A couple of days ago the Miami-Herald “broke” a news story that claimed Florida changed its COVID-19 data recently – something that I had worked with a Sun-Sentinel reporter to break 18 days prior. Now the story has been picked up by MSNBC, the Wall Street Journal, and other national news services. However, none of these stories have addressed the full impact of Florida changing COVID-19 cases and deaths from date of reporting to actual date of occurrence.
1. While technically correct the new FL method is only used by a few other states to report their deaths to the CDC. Contrary to the WSJ article’s claim, TX is not using the FL method – or if TX is using the new method they only have a 1-2 day lag in recording deaths. The figures below show how CDC currently stores and displays FL and TX death counts. As you can see FL death counts show a peak two weeks ago and a drop-off to near zero today. While TX death count curve shows a steep rise to near-record highs. The FL data for the last 4 weeks will be continually adjusted upward over the next 4 weeks – delaying accurate information for 4 weeks.
2. Allowing some states to report one way and other states to report a different way is a basic Data Science 101 error by the CDC. You add apples to oranges and you get a meaningless fruit salad. The CDC according to the WSJ has conceded that some states report the FL way and others such as TX report the traditional way.
3. Some states like CA are using actual dates reporting but because they update their data daily, databases and modelers can choose to use either time series in their models. For FL, though, the wrong death count could be in the system for 7 days (eg. JHU still shows 43,979 as the total cumulative death that was reached and reported 6 days ago for FL).
4. If the CDC adds all the states up for the USA total, it distorts the picture of USA death counts and confuses the public and data scientists trying to forecast and advise POTUS. FL accounts for ~20% of the US total deaths. The new FL method understates deaths for the last 3-4 weeks which are used by many models to predict future death trends causing them to understate their forecast and causing POTUS to react too late.
Just as good military intelligence is essential for the conduct of a successful war – accurate, timely, and consistent information is essential for successfully winning the war on SARS-CoV-2. USA intelligence on the COVID-19 pandemic had been poor and has not improved much this year. Inadequate testing resulting in very high positivity rates, too little genomic testing to understand Delta’s contagiousness and ferocity, dropping breakthrough tracking just when vaccine efficacy may be waning, and switching reporting methodology and allowing states to drop daily reporting in the middle of a pandemic are just a few examples that could cause the USA to lose the war. The CDC should set the standards for consistent daily reporting and help all states to achieve this. Viruses do not respect state boundaries and the CDC should not let states dictate what information states give them. When COVID19 cases are doubling as quickly as every 3 days in some jurisdictions earlier this summer, cases can grow a thousandfold in a month and quickly swamp hospitals causing unnecessary deaths. Timely, accurate, and consistent information is key to making the right decisions to mitigate the spread and win the war.