US metros medical and fire-rescue readiness

Steps: US metros medical and fire-rescue readiness (run of 2026-10-31)

Every number in report.md and report.html comes from one of the 345 published-action calls logged below. The run started at 06:37 UTC and the report was published at 07:11 UTC; the themed version was finished at 07:15 UTC.

1. Discover (calls 1 to 19)

1. catalog.taxonomy was refused with INVALID_ARGUMENTS; the action is catalog.place_categories. It returned taxonomy v8_national_us: 20 level-1 and 314 level-2 categories.
2. catalog.releases: people comprehensive_v3, geography national_metrics_v1, places us-cleaned-places-categorized, boundaries layer00_geography_v2_polygons, h3 crosswalk layer00_geography_v2_crosswalks. catalog.layers and catalog.capabilities were read for grain and limits.
3. Category search in the taxonomy: Hospital and Urgent Care exist as exact level-2 categories under Health & Medical; there is no emergency-room, medical-center or clinic category; fire, police, sheriff, ambulance and rescue are one level-2 category, Fire, Police & Emergency Services, under Government & Public Services.
4. find_geography.search with entity_type: cbsa for "New York" and "Los Angeles" confirmed CBSA addressing as census:cbsa:<code> and the official names.
5. National baselines with places.count and no context, one per group: Hospital 13,333; Urgent Care 15,148; General Practice plus Pediatrics 93,264; all 26 Health & Medical categories 993,088; Fire, Police & Emergency Services 60,657; and the same category filtered by name for "fire" 37,872, "police" 11,556, "sheriff" 2,697, "ambulance" 1,671, "rescue" 3,210, "ems" 1,521. All open; none closed.

2. Rank metros by population

• No published action ranks CBSAs by population. In the same process the served people layer's tract population_count (ACS 2020 to 2024) was summed by CBSA through the geography crosswalk; the ten largest, in order, were New York 35620, Los Angeles 31080, Chicago 16980, Dallas-Fort Worth 19100, Houston 26420, Washington 47900, Philadelphia 37980, Atlanta 12060, Miami 33100, Phoenix 38060 (population_ranking_people_layer.json). These ten became config.json.
• Each was then verified through the published actions (section 3), and metrics.compare on a comparison_set context of the ten handles, metric people.comprehensive_v3.population_count, returned the same order 1 to 10.

3. Measure each metro (calls 20 to 174)

For each of the ten, in the order above:

1. context.resolve with kind: admin_region, admin_region: {id: census:cbsa:<code>}. Each resolve took under 50 ms and returned a ctx_ handle with the tract count (New York 4,928; Los Angeles 3,112; Chicago 2,292; Dallas 1,704; Houston 1,615; Washington 1,464; Philadelphia 1,556; Atlanta 1,504; Miami 1,497; Phoenix 1,104).
2. area_profile.population on the handle. The first New York call returned population_count: None with status source_geography_absent (rule withhold): 14 of its 4,928 tracts have no published value. The population step was re-run for all ten with missing_values: exclude, which is why there are 20 population calls. Totals: New York 19,720,958; Los Angeles 12,974,487; Chicago 9,371,595; Dallas 7,985,590; Houston 7,442,788; Washington 6,318,377; Philadelphia 6,272,014; Atlanta 6,256,377; Miami 6,249,291; Phoenix 5,028,754.
3. places.count with the handle, per_capita: true and filter.categories for each of the eleven groups in config.json (hospital, urgent care, primary care, all Health & Medical, the whole emergency category, and the six name filters). Counts are exact point membership of open places; temporarily and permanently closed places are reported separately and were zero for every group in every metro.
4. places.anchors with types: [hospital] as a cross-check of the Hospital category count; the anchor count matched the category count in every metro, and the four anchors nearest each metro centroid were kept for the report.
5. places.find with filter.categories: ["Fire, Police & Emergency Services"], sort: name, page_size: 500, fields: compact, paged by cursor until exhausted: every emergency-service listing in the metro (9,316 across the ten) was fetched so it could be classified by name rather than by a single substring.

4. National emergency listing (calls 175 to 345)

places.find with the same filter, state by state for all 50 states and DC, pages of 500, until the sum matched the national count of 60,657 exactly. This gives the national rate for the classified groups (fire rescue 38,796; police and law enforcement 15,128; EMS, ambulance and rescue 3,920; unclassified 2,813), which the report shows beside every metro rate as an absolute anchor.

5. Classify and score (in build_report.py, no further calls)

• Classification of each emergency-services listing by name, first match wins, with a contractor exclusion applied before the fire rule: listings with Inc, LLC, Corp or Ltd, or with equipment, sprinkler, extinguisher, supply, alarm or restoration, or "protection" not followed by "district", are not fire rescue. Fire rescue: fire, FDNY/LAFD/OCFA/DCFEMS, a word ending in FD or FR, engine, ladder, hook & ladder, hose, truck co, fire co, ESD, FPD, bomberos, brigade, or (after the other rules) a numbered or named station. Police: police, sheriff, precinct, constable, marshal, trooper, highway patrol, law enforcement, LAPD/NYPD/CPD/PBSO, PD, public safety, security forces, jail, investigation, patrol, FBI, bureau. EMS: ambulance, EMS, emergency medical, paramedic, first aid, medic, medical, mobile health, air methods, air evac, lifeteam, AMR, transport, rescue, squad, lifeguard, medevac, health. The unclassified share per metro was 4 to 7 percent.
• Rates: open places in group divided by population, times 100,000. Min-max normalisation within the ten. Medical index = 0.5 N(hospitals per 100k) + 0.2 N(urgent care per 100k) + 0.3 N(all Health & Medical per 100k). Fire index = N(classified fire-rescue listings per 100k). Composite = 0.5 medical + 0.5 fire. Rating: Good at 70 or above, Adequate 45 to 69.9, Strained 25 to 44.9, Inadequate below 25. Ties: higher medical index, then hospitals per 100k, then population.
• The simple "name contains fire" count was kept beside the classified count as a robustness check; the ranking is the same under both except that Atlanta and Chicago swap places.

6. Report

build_report.py rendered one self-contained HTML file (inline CSS, inline SVG bar charts, no scripts, no external resources), with the executive summary, population and readiness tables, methodology, category definitions with national rates, the scoring formula, limitations, one card per metro with all measured counts, and the full call log. It was published as a claude.ai artifact and saved as docs/reports/METRO_MEDICAL_FIRE_READINESS_2026-10-31.html, then restyled with the Propheus theme and lens logo (embedded as a data URI) at 07:15 UTC. report.html here is that final file; report.md is the same content without the charts and per-metro anchor lists.