Update pharmacy locations for ITA and HUN

[cjacobscrisioni]

OECD just ran checks on the official pharmacy location data for a number of countries. See email 2026-09-07. Italy and Hungary have been adjusted (I believe geolocated by OECD themselves). See attached locations zip file and quality check document:

pharmacy_quality_checks_07_Sept_2026.docx

New locations for Hungary and Italy need to be included. The file:
hun_ita_checked_existing_pharmacy_locations_20260907.zip

[MaartenHilferink]

Installed and wired in (29fcf7f); the affected areas are being rebuilt and re-swept

What the delivery contains

Both files are the standard lat/lon parquet (id, name, ref_date, lon, lat, geom; Italy also address), WGS84, no missing coordinates.

	rows	replaces	source per the OECD quality-check document
ita_pharmacies.parquet	20,635	ESPON 2021 point shapefile, 12,991	Ministry of Health list (20,652 active, 20,794 unique incl. 142 closed, 1,800 without coordinates); official coordinates were duplicated/implausible, so all were geolocated by the OECD (Mapbox > ArcGIS > manual; 28 of 30 rural checks correct)
hun_pharmacies.parquet	3,094	— (new country)	National Center for Public Health and Pharmacy; coordinate-less source geocoded by the OECD; records starting with "DR." removed (accuracy 76 % → 90 %)

Installed under SourceData/NetworkModel_EU/Locations/Pharmacies/, with the June unchecked Italy parquet kept as ita_pharmacies_20260622_unchecked.parquet and the OECD document and zip beside them for provenance.

Config

cfg/main/SourceData/Locations.dms: Italy reads the parquet again through CountryDataT (the shapefile template stays for the next such delivery); Hungary added as country code hun in the enum and in AvailableCountry (19). Hungary's TomTom network was already on the source share.

Why this is a rebuild, not just a re-sweep

The cached network (FinalSet) carries the connector links from the pharmacy points to the road nodes, so it depends on the pharmacy set: the descriptives run on the new Italy data failed inside the cached ITC network with DestNode_rel: out of range. So for Italy's five NUTS-1 areas and for Hungary the network and both ODs are rebuilt (run_rebuild_all.ps1 -Areas ITC,ITF,ITG,ITH,ITI,Hungary), then the descriptives for Italy, Hungary and the five, then full sweeps (both cost functions) for the six. Italy's candidate set changes as well (candidates = ≥50-pop cells ∪ pharmacy cells).

Italy goes from 12,991 to 20,635 pharmacies, so its baseline (more locations, less travel), its descriptives (residents per pharmacy 4,520 → about 2,850) and every Italian frontier move; the earlier ITF/ITG rankings on the improvement lists are superseded. Hungary is a new row everywhere: descriptives, region slide, S1/S2 tables, the point clouds and ranked lists, and the aggregate (41 → 42 disjoint areas). Its policy-typology class is still needed for the coloured charts — Chris, is Hungary in the typology deck?

Results follow here when the sweeps finish (the runs are combined with the #52 S1/S2 refinement).

[MaartenHilferink]

Hungary's typology is on slide 13 of the typology deck (Type 1 Rule-of-law-anchored, HIGH formalisation: quota + distance + dispensing physicians), so no input needed - added in 8963d27; ignore the question above.

[MaartenHilferink]

Results (216ac77): the south of Italy was three pharmacies in five short — the Italian "improvement potential" was missing data

Italy's five NUTS-1 areas and Hungary were rebuilt from the network up (the cached network carries the pharmacy connector links) and fully swept on both cost functions; Finland came in alongside (#44). Deck doc/lambda_sweep5.pptx (76 slides), README and the #45/#48 exports are regenerated.

What the OECD list changed

area	ESPON	OECD	residents / pharmacy
Italy	12,991	20,621 in the study area	4,520 → 2,848
ITC Nord-Ovest	4,463	5,624	3,548 → 2,815
ITF Sud	1,951	4,955	6,831 → 2,690
ITG Isole	1,449	2,332	4,352 → 2,704
ITH Nord-Est	2,786	3,805	4,138 → 3,030
ITI Centro	2,342	3,919	4,996 → 2,986
Hungary	—	3,094	3,119

ITG's baseline mean travel goes from 4.9 min (after the landbody fix) to 2.0 min; ITF from 6.7 to 2.0 min.

Consequence for the rankings

Relative improvement potential (rectangle / baseline count × travel), LINEAR:

rank	ESPON list	OECD list
1	ITG 0.384	SE1 0.117
2	ITF 0.268	Norway 0.112
3	ITI 0.129	Lithuania 0.086
4	ITC 0.128	SE2 0.085
6–7		ITG 0.073, ITF 0.071

Under LOGISTIC the five Italian areas are in the bottom ten of 43. The sparse Nordic and Baltic areas lead now, which is also where the 5-nearest choice set binds hardest (Norway: 18.6 % of residents one closure from stranding at S2; ITF on the new list 1.8 %, was "a third of the population"). The earlier decks' headline — southern Italy far above the field — was the ESPON list, not the geography. The scope slide (p8) and the roadmap (p65) say so.

Aggregate over the 43 disjoint areas (Hungary and Finland included): 51,350 pharmacy cells today; S1 −23 % travel at today's count, S2 −30 % locations at today's travel (LINEAR; were −26 % / −34 % over 41 areas) — Italy on its full list sits closer to its frontier.

Hungary

3,094 pharmacies in 1,865 cells, 9.65 M residents, baseline mean travel 3.6 min, no unreachable cell; S1 −19.5 % travel, S2 −25.8 % locations; ranks 20th of 43 (LINEAR). Typology Type 1 / HIGH from slide 13 of the typology deck.

Provenance: the OECD document and zip are under SourceData/NetworkModel_EU/Locations/Pharmacies/ next to the installed parquets; the June unchecked Italy file is kept as ita_pharmacies_20260622_unchecked.parquet.