Evaluating Alabama’s August 11th Daily Death Count
On August 11th ADPH reported a daily death total of 48. This is the second highest daily total on record. As always a dive into the data reveals there is more to the story than the headlines suggest.
Fig. 1 shows the deaths reported by week for both ADPH and the CDC. The CDC only reports deaths by week so at this time a daily comparison is not possible. The CDC data I am using can be found here. Note this data is updated each weekday so reports pulled on August 12th and later will show different totals. An important item to note is the CDC records deaths by date of death while the ADPH reports the death as of the date reported. For example, if a person dies on August 1st and the death certificate is added to the ADPH system on August 10th the CDC will add it to the total for the week ending August 1st while ADPH will include the death in the week ending August 15th. This is an important distinction. To accurately assess patterns we have to know the actual dates of death. Grouping deaths from multiple time periods into a single day distorts the data and makes any meaningful analysis impossible. Analyzing trends by reported date will tell you the pattern reported deaths follow. I suppose that could be useful to evaluate how the system is functioning but is worthless when evaluating what is happening in the real world.
As you can see from Fig. 1 as of this writing ADPH has reported 8 more deaths than the CDC. This makes logical sense since there will be some lag in getting the data to the CDC. However, the numbers are close enough we can start doing some analysis using basic logic. The first thing to notice is the CDC typically reports more deaths than ADPH. Again, this makes sense. The CDC pushes deaths back to the date of death so while ADPH played catch up the CDC has since gone back and allocated the deaths to the correct week. As deaths rose this problem was amplified. For the week ending July 25th the total under count for Alabama was 278. What does this mean? Simply put, of the 368 deaths reported by ADPH between July 26th and August 11th 278 or 75.5% actually occurred on July 25th or earlier. This massively skews the data and makes it all but impossible to accurately chart whether deaths are trending up, down, or are staying flat. Without context the daily numbers are only useful for generating headlines, not providing meaningful information.
The second thing to note is deaths truly are rising. As noted in a previous post Alabama case counts truly are rising. With it deaths will also grow. Again August will be interesting as Alabama is nearing herd immunity and cases should begin dropping with a drop in deaths expected to follow. A cursory look at the data indicates the weeks ending August 8th or 15th may be the peaks for death when all is said and done.
A third observation is more forward looking. It seems logical that if ADPH daily death numbers lag then when AL deaths do peak and start decreasing it will not be immediately apparent. If roughly a two week reporting lag is used based on the August 11th reports then it seems likely deaths will have been declining for 14+ days before it is noted by ADPH. That delay is not acceptable when public policy decisions are being made based on this data.
Figure 2 shows the same data in figure 1 in graph form. As you can see ADPH has generally been below the CDC reporting each week. It is not much on a weekly basis but it comes to a head in August when the CDC has yet to report any deaths after August 1st and ADPH has reported 228. The totals of the two datasets are only 8 apart. This means the ADPH data from August 2nd through August 15th on the graph truly is catch up from March through August 1st. The red CDC line is useful. We can see deaths are trending up which makes sense given the increasing cases. The sharp drop at August 1st will grow as new deaths are added and backdated to date of death but for now I believe the graph does a good job of showing the disparity between what is reported and what actually happened.
The bottom line is ADPH data is not that useful for identifying trends in the data. Without the CDC database we would be totally in the dark about what is happening in the state. For now it seems the best option is to ignore any press releases or news items that do not provide this information with proper context.
