Day 10 · Prompting only — One sentence, one health map

Data

BD Hospital Dataset (hospitals, clinics, pharmacies), fire stations, garment factories, economic zones and major roads from the GIS Data folder; BBS Population and Housing Census 2022 thana populations; city-corporation boundaries. The registry is incomplete and urban-biased; population per 250 m cell is a dasymetric model; distances are straight lines in UTM 46N.

Methods

1. Study areas: the two Dhaka city corporations (DNCC + DSCC, 53 thana polygons, 10,250,630 people) and Chattogram City Corporation (15 thanas, 3,273,329 people), BBS Population and Housing Census 2022. White gaps on the maps are outside the corporations (cantonment, airport, water).
2. Health facilities: the BD Hospital Dataset in the GIS Data folder, split into hospitals, clinics and pharmacies from the type and name keywords; pharmacies within 25 m of each other merged; an ‘emergency / 24 h’ flag from the name and type. Dhaka: 475 hospitals, 507 clinics, 2,049 pharmacies, 44 fire stations. Chattogram: 160, 51, 405 and 24.
3. Where people live: the census gives thana totals only, so each thana total is spread over a 250 m grid in proportion to a Gaussian kernel (σ = 400 m) of settlement proxies — mobile-phone towers, mosques, schools, health points, pharmacies, hospitals — plus a small floor. Cell sums equal the census total exactly (checked).
4. The prompt: written first and kept verbatim (shown on the map). It names the data (BD hospital dataset, census), the unit (250 m cell), the measure (distance to nearest hospital, clinic, pharmacy), a threshold (10 minutes) and the outputs. Every vague word in a first draft (‘health’, ‘access’) was replaced by a number or a layer.
5. Coverage curves: people-weighted share within 100–3,000 m for hospital, any care, pharmacy and emergency care; thana tables with median distance, % within 500/800/1,000/2,000 m, people per hospital, care points per 100,000.