USDR Appointment Availability Data - Coverage

• In this report we take cursory look at slot-level COVID appointment availability data as gathered by USDR’s UNIVAF project.
• We look at the three weeks between 2021-05-24 and 2021-06-14
• We count a slot as “booked” if we see it available first, and then not anymore (up to two hours before the appointment starts).

TLDR

• Slot-level Appointment data availability ranges from 19% (NH) to 95% (CO) of loctions.
• We estimate that about 20% of vaccines administered in the last two weeks are covered in our slot-level data. Coverage is the highest in soutern states.
• For locations that report slots and/or appointment numbers, the reporting is very stable.
• For states, ME, VT, WI have the most unreliable reporting. For providers, walgreens has the most unreliable reporting.

Sample Data

Here are the actual availability counts (not necessarily slot-level data) for a sample of 25 locations. You can see different supply patterns, and some example locations that only publish “has availability” (rather than “x appointments”) info.

Slot-level data over time

Coverage by state

We look at how many locations we have, how many have availability at all, how many report slot-level data. Then we compare these to the population and the number of vaccinations that happened during this time (two weeks). These numbers might not match up (slots could stop being available without being booked, not all bookings will be of a 2nd shot, vaccinations could be reported later), so the “share of vaccinations covered” statistic could go over 100%.

Coverage of locations compared to vaccinatethestates.org

Here is the total number of locations by county, in our database, as compared to vaccinatethestates.org. This does not take into account whether the location has availability or slot data.

As you can see, VtS often has somewhat more locations than we do.

Here is the distribution of the ratio of the counts:

VtS has more locations almost everywhere, except for DE and NH

Continuity of data

Here we look at how stable data updates are. We count the minutes between updates for each location (taking into account that I’m scraping our API every hour), and take the count, number of updates missed (delta > 90 minutes), average minutes, median, 95 percentile, and 99 percentiles. Then we average those by state and provider.

By state:

By provider (with at least 100 locations):