Bayesian spatiotemporal small-area analysis · Texas DSHS records, December 2020 to May 2023
COVID-19 vaccination coverage across Texas counties
How first-dose coverage of residents aged 16 and over changed in each county, where the gaps between counties opened and widened, and how uptake varied with age composition, household income and recent cases. Every number describes vaccine receipt as recorded in the state registry, not willingness.
Rscript --vanilla R/16_build_site.R.County map
Choose what to map, move through time, and select a county to see its record. Hover or tap a county for its value; hold Shift and click a second county to compare the two. Zoom with the buttons, a double-click, Ctrl and the mouse wheel, or two fingers on a touch screen; then drag to move. The charts further down show their values the same way: hover, or tap and slide sideways.
Show all counties as a table
The gaps between counties widened, and most of the gap opened early
Show as table
- Across those counties the distance between the 90th- and 10th-percentile county was 9.9 points at the end of January 2021, 17.6 at the end of February and 22.1 on 24 May 2021. It ended the study at 26.3 points.
- The size of the gap was set early, but not yet which counties were on which side of it: county coverage at the end of February 2021 has a rank correlation of 0.37 with coverage in May 2023, against 0.90 for coverage on 24 May 2021.
- After May 2021 the gap kept widening because uptake among adults still unvaccinated was itself higher where coverage was already higher: 1.49 times higher per 10 points of starting coverage (95% confidence interval 1.42 to 1.56).
Uptake among adults not yet vaccinated
Show as table
- Reporting peaked in the first two weeks of April 2021, just after eligibility opened to all adults: 26.5% of still-unvaccinated adults per 30 days statewide.
- After the baseline the rate fell to 6.0% in early summer 2021, rose to 12.6% during the Delta wave, and was at or below 1% from June 2022.
- The median county ran well below the pooled rate throughout, because the most populous counties vaccinated faster.
Age composition, income and recent cases
Hazard ratios per standard deviation of each county characteristic, by period. A ratio above 1 means higher uptake among adults not yet vaccinated in counties with more of that characteristic.
Show as table
- Age. While eligibility was restricted, counties with more adults aged 65+ were not behind (1.04 (0.97 to 1.11)). From April 2021, when all adults became eligible, the ratio is 0.89 (0.83 to 0.96), and over the main study period 0.83 (0.78 to 0.88).
- Recent cases. Counties with more reported cases in the previous 30 days had higher uptake in every period (1.06 (1.03 to 1.09) over the main period).
- Income. Associated with coverage reached by May 2021 (odds ratio 1.05 (1.01 to 1.10)); for later uptake the ratio is 1.05 (0.98 to 1.13), an interval that includes no association, and it is the least stable of the three across alternative analyses.
- These are associations between county characteristics and county outcomes. They do not describe individuals.
Look at the counties yourself
Plot any two county measures against each other. Each dot is a county; the selected county is marked.
The first five months rest on an assumption
Until mid-April 2021 a share of recipients had no county on record. On 15 April 2021 the state assigned about 664,000 of them to counties in one edition, very unevenly. The reconstruction carries each county's share back to earlier dates, scaled to the number of unassigned recipients published at the time.
- The assumption predicts how published county counts should have grown whenever the unassigned total rose or fell. Over six two-week windows the predicted and estimated effects correlate at 0.97 with counties weighted by size and 0.46 with counties weighted equally: well supported for large counties, less for small ones.
- With no allocation at all the conclusions hold; individual counties' coverage moves by up to 8.5 points.
- 25 modelled counties whose early records kept changing are left out before 24 May 2021 (grey on the map at early dates).
- Who was ahead changed when eligibility opened: relative uptake to 31 March 2021 correlates 0.28 with relative uptake after 24 May 2021, against 0.77 for April-May 2021.
How well the models predict
Show as table
- Spatial structure lowered the error for counties with no data by 14% (random counties withheld) and 13% (whole regions).
- County-specific trajectories lowered the error for a county's next interval by 8% and made later intervals better calibrated, but made predictions for counties with no data 8% worse.
- No model forecast the statewide level of the next interval well.
- Smoothing does not hide local differences: for 136 of 238 counties the county effect differs from its neighbours' with 95% posterior probability, and a prediction from neighbours and predictors alone would miss a county's May 2023 coverage by 3.6 points on average (map: "Neighbours-only prediction error").
Method and cautions
- Outcome. Residents aged 16+ recorded with at least one dose, by county of residence, from archived Texas DSHS workbooks. The records support 16+, not 18+, and the dates on which editions survive, not calendar months.
- Denominator. Census Bureau Vintage 2025 estimate of residents aged 16+ on 1 July 2021, held fixed.
- Model. Binomial model for first doses among adults not yet vaccinated, complementary log-log link with interval length as offset, BYM2 county effect, shared time effect, county-specific mean-reverting trajectories, fitted with R-INLA.
- Sixteen counties are not modelled because their recorded count of vaccinated adults exceeds their adult population at some point; most are along the Mexican border. They are grey on most maps.
- Recorded coverage is a complete count. Its uncertainty comes from the records and the population estimates, which the credible intervals do not cover. Uptake rates in high-coverage counties are the least reliable numbers here.
- Dates are reporting dates. Early in the rollout reporting lagged vaccination by days to weeks.
What the terms mean
- Coverage
- Residents aged 16+ recorded with at least one dose, as a share of the county's adult population estimate.
- Uptake per 30 days
- The share of adults not yet vaccinated at the start of an interval who received a first dose, scaled to 30 days so that intervals of different length compare.
- Relative uptake
- A county's uptake rate divided by that of the average county in the same intervals. 2 means twice the average rate.
- Hazard ratio
- How many times higher the uptake rate is in counties one standard deviation higher on a characteristic. 1 means no association.
- 95% credible interval
- The range that holds the modelled quantity with 95% probability, given the model and the records. It does not cover errors in the records or the population estimates.
- Spatially smooth effect
- The part of a county's uptake level that it shares with neighbouring counties and that the three predictors do not explain.
- CRPS
- Continuous ranked probability score: the distance between a prediction, with its uncertainty, and what was recorded. Lower is better.
The written reports, code and every result table are in the project folder:
docs/00_summary.md, docs/01_source_inventory.md, docs/02_report.md,
docs/03_early_rollout.md.
This page is built from the result tables by R/16_build_site.R.