Top news, reports and insights for today:
- Daily headline summaries for Monday:
- California public health officials have notified 180 people, who attended a religious gathering on Mother’s day in violation of the stay-at-home order in Butte County, that they may have been exposed after at least 1 person tested positive for coronavirus (ABC News)
- Anonymized cell phone data from 5 states shows lockdown protestors travelled widely within and between states, raising the possibility that such protests are playing a role in disease transmission (The Guardian).
- Moderna announced “promising results” from initial human trials of its vaccine. This is misleading because phase I trials do not address the efficacy of a vaccine (ability to prevent infection). These tiny studies first address safety and whether antibody response is plausible. Many candidate vaccines that show evidence of safety fail to demonstrate efficacy in definitive phase III trials (Washington Post)
- Using 1-week growth factor to track the speed of the epidemic
How fast or slow are cases growing? That’s a basic question. There are several questions we can ask our data. Here is a new one (technically not a new method, but a new visualization). It’s important to pay closest attention to growth factors in an epidemic. That is the ratio of new cases between two time periods. This comes in a variety of flavors, but let’s zoom in on a simple one. What is the ratio of cases last week to cases in the week before that? That ratio gives a 1-week growth factor (graph below). Values >1.0 indicate new cases are growing. Below 1, the rate of new cases is slowing. Of course the number of cases can’t decline, but the rate of cases can.
Let’s start with the Northeast; all states except Maine are seeing declining rates of case increase. That’s good news. Maine’s high GF is partially explained by an administrative change 2 weeks ago that resulted in a negative case growth after probable and confirmed cases were separated. The Midwest was a mixed bag, with 3 states rising (Michigan, North Dakota and Oklahoma) and 6 states falling (Indiana, Kansas, Missouri, Nebraska, Ohio and South Dakota). In the West, new case growth slowed in California, Colorado, Oregon, Washington and Wyoming. Growth in new cases was seen in Arizona, Idaho, and Nevada. Growth factor estimates were unstable due to small numbers in Alaska, Hawaii and Montana. New case growth was seen most consistently in the South, were increases were seen in 9 of 13 states. New case growth slowed in Alabama, Kentucky, Mississippi and West Virginia, but grew fastest in Tennessee, Arkansas, Louisiana, Texas and North Carolina (in that order).
What this means: No measure of the epidemic’s speed and severity is perfect, but consecutive-week growth factors are a useful tool that summarizes recent trends. The areas of greatest concern now are in the Midwest and South, where accelerating case growth can be seen in 13 states. Particularly worrisome are increasing case growth in Michigan and Louisiana, two states that have been especially hard hit and had been trending in a more positive direction.
- I still have Georgia on my mind
All eyes are on Georgia. That state has been under the microscope since being one of the last to issue stay-at-home orders (April 3) and first to reopen (April 24). If we are going to see a surge in cases resulting from reopening, we expect to see it first here. Multiple news sources indicate that such a surge has not yet appeared. Some in the state are claiming an early victory. Experts caution that we may not see the resurgence in infections for 4-6 weeks (see headlines from yesterday). I thought I would take a closer look at what is going on in that state.
Let’s start with the basics. The top graph below is from today’s Atlanta Journal-Constitution (AJP) website and shows daily cases and a 7-day rolling average. Since reopening began on 4/24 there has been a declining trend in cases. But have we waited long enough?
Thanks to the lack of national standards, every state does testing differently. In Georgia, the reporting lag may be greater than other states for a host of reasons. Recent accounts in the AJC show a track record of embarrassing gaffs and sloppy counting in the state’s reporting system. Beyond that, like many states, Georgia backdates cases to when the patient first reported symptoms. This means the most recent 2-3 weeks of data is incomplete and will always show a downward trend. For example, as the AJC points out, only 21 new cases were initially reported for May 1 on the GDPH website. Today, it shows 1,115 new cases attributed to May 1. I’m sure this sounds like a nerdy detail, but its important. Since new case reports are backfilled to the date of reported symptom onset, all new case reports for at least 2 weeks should be regarded as provisional and not suitable for establishing trends. Add to this, the 2-3 weeks it takes for a given case to go from infection to symptoms to seeking treatment for those symptoms, and we have to extend the cone of uncertainty even further. So how far back do we have to go before we can even believe the numbers in Georgia? No body knows for sure, but we can’t conclude that cases are dropping since the reopening. That is conclusion #1.
To investigate this issue further, I made my own graph using data from the Covid Tracking Project. Instead of new cases this shows total tests done per day since April 1 (blue bars) along with the percentage of positive tests (aka the test positivity rate). I made this graph to see if the apparent flattening of cases since reopening is related to a change in the number of tests being run. Based on the above, there is no way to tell because the date a test is reported is not related to the date it was run. There are at least 2 data anomalies that confirm what AJC has said about the Georgia data system. For some reason, 100% of the tests reported on May 3 were positive. More surprisingly, a record 30,000 test results appear in the system yesterday. The number of tests has risen over the last 6 days. I’ve never seen tests double on Sunday, a day almost uniformly associated with a weekend slow-down. This trend is coupled with a strong decline in the TPR over the same period. Georgia was averaging under 5,000 tests a day until the it reopened on April 24. Since then, the number of tests has been rising linearly, doubling to 10,000 a day by May 11, then, doubling twice in the last 4 days. There is no explanation for this trend. I can’t reconcile these opposing patterns. The backfilling of tests conflicts with the significant rise in tests.
The Bottom line: It may be weeks before we know for sure what the impact of reopening is in Georgia. The lack of an uptick in cases is to be expected given the data lag. What we do know is that all eyes are on this state. Public Health officials are under tremendous scrutiny and the governor’s job is on the line. Recent polls say his approval rating is among the few that has gone down during the pandemic. It is inarguable that powerful incentives exist for Georgia to hide cases. Time will tell if our confidence in the data from this and other states rises or falls.