AXIOM|
Layer logoLayerOverwatch logoOverwatchLocus logoLocusCodex logoCodexRouteShift logoRouteShift
Docs
Analysis

449,000 CDC Health-Indicator Events Landed in 3 Days Across 288,532 Census Tracts. Zero Have an H3 Index. Cell Scores Has No Column to Put Them In.

449,000 census-tract health records landed in three days. None of them carry an h3 index, and cell_scores has no field built to receive them even if they did.

The Setup

449,000 CDC PLACES health-indicator events landed in Locus's axiom_events table over the trailing three days, spanning 2,932 counties and 288,532 distinct census tracts — cholesterol screening rates, checkup rates, high blood pressure prevalence, obesity, disability, and eight other measures, each carrying a uniform 0.95 confidence score. Zero of those 449,000 rows carry an h3_index.

The Chain

Every other signal that reaches Locus's composite cell score gets there the same way: a row lands with an h3_index, and the scorer joins on it. Building permits carry one. Civic records carry one. This batch doesn't. The five busiest measures alone — cholesterol screening (74,640 rows), general checkups (74,559), high blood pressure (69,940), obesity (65,823), and disability (59,079) — account for the bulk of the 449,000, each one tagged to a census tract in location_name and a county_fips in the metadata, but not to a single hex cell. cell_scores doesn't have anywhere to put this data even once the join key exists: the table's signal columns cover business vitality, demographics, safety, accessibility, development pipeline, and a handful of others, with no dedicated health field and nothing in signal_scores currently populated from a CDC source. The data arrived faster and wider than anything else added to the corpus recently — 288,532 tracts in three days is closer to a national backfill than an incremental load — and it arrived with no path into the system that would act on it.

The Implication

A census tract centroid converts to an h3 cell cheaply and deterministically; that isn't the hard part. The hard part is that even a clean join wouldn't currently change a single cell's composite score, because nothing in the scoring pipeline reads a health signal. Two gaps are stacked here, not one: the join key is missing, and the destination for the joined value doesn't exist yet either. Until both are addressed, 449,000 rows of chronic-disease and preventive-care data — arguably one of the more decision-relevant civic signals for long-horizon CRE and public-health-adjacent underwriting — sit in the table, queryable directly, invisible to every product surface that reads cell_scores.

What to Watch

Whether ingestion volume keeps climbing at this rate over the next few days — suggesting an active, possibly ongoing national load — or whether the 449,000-in-three-days figure was a single backfill batch that won't repeat. That distinction changes how urgent the h3 backfill is.

Limitations

A flat 0.95 confidence across every measure and every tract reflects trust in the CDC PLACES source itself, not statistical reliability at the tract level — some of the 288,532 tracts carry total_population counts small enough that a headline rate should be read with caution regardless of source confidence. This snapshot only confirms h3_index is null for the trailing three days of ingestion; it doesn't establish whether that's true of the source's full historical load or specific to this window.

--- Data as of 2026-08-28. Source: Locus `axiom_events`, event_type='health.indicator', source='cdc_places', trailing 3 days (449,000 rows, 2,932 counties, 288,532 distinct tracts); h3_index null-check on same window; `cell_scores` schema review for health-related columns.

cdc-placeshealth-indicatorh3-hex-cellsdata-qualitycell-scoresscoring-coverage-gap

Location intelligence derived from 85 catalog feeds across 22 metro markets. Scores updated continuously.

← Back to Intelligence