NYC Environment & Health Data
The indicators published on the NYC Environment & Health Data Portal, run by the NYC Department of Health and Mental Hygiene (DOHMH): heat, air quality, asthma, housing, pests, water, and more, by neighborhood, borough, and time period.
Portal: https://a816-dohbesp.nyc.gov/IndicatorPublic/
Source repository: nychealth/EHDP-data, branch production. This is the data the portal itself loads.
License: Apache-2.0 (NYC DOHMH, nychealth/EHDP-data)
Refresh: quarterly. Each scrape reads one commit of production and logs its hash.
Imported by Knowledge Futures. Not maintained or endorsed by the City of New York.
Record types
| Type | Count | What it is |
|---|---|---|
Indicator |
267 | A topic, e.g. "Heat stress: yearly emergency department visits" |
Measure |
731 | One way of quantifying an indicator: a count, a crude rate, an age-adjusted rate, a percent, a score |
TimePeriod |
307 | A year, quarter, season, multi-year span, and so on |
Geography |
835 | A place: citywide, a borough, or a neighborhood in one of several neighborhood systems |
Comparison |
75 | A named group of measures the portal charts together |
Observation |
532,404 | One value: one measure, one geography, one time period |
How they reference each other (all references are record ids inside this collection):
Observation.measureId -> Measure Measure.indicatorId -> Indicator
Observation.indicatorId -> Indicator Measure.availableTimePeriodIds -> TimePeriod
Observation.geographyId -> Geography Measure.linkedMeasureIds -> Measure
Observation.timePeriodId -> TimePeriod Indicator.measureIds -> Measure
Indicator.comparisonIds -> Comparison
Comparison.indicatorIds / measureIds / geographyIds -> Indicator / Measure / Geography
Ids
Indicator:indicator:<IndicatorID>, e.g.indicator:2075Measure:measure:<MeasureID>, e.g.measure:537TimePeriod:timeperiod:<TimePeriodID>, e.g.timeperiod:49Comparison:comparison:<ComparisonID>Geography:<GeoType>:<GeoID>, EHDP's own strings, e.g.UHF42:101,CD:101,NTA2020:50101,Borough:1,Citywide:1Observation:<MeasureID>:<GeoType>:<GeoID>:<TimePeriodID>, e.g.537:UHF42:101:49
The numeric source ids are also kept as fields (indicatorId, measureId, timePeriodId, comparisonId, geoType plus geoId), so you can match them to portal URLs and the source files.
Observation fields
value: the number, unrounded, as the source stores it. Absent when the source suppresses the value.ciLower,ciUpper: 95% confidence interval, where the source gives one (survey estimates mostly; about 29,000 observations).displayValue: what the portal shows, rounded and annotated, e.g.11.1*,**,4,000^^.flag: the annotation symbol from the display value.*small numbers, interpret with caution;**or***suppressed;^^large relative standard error or coefficient of variation;^outliers excluded;†not shown (e.g. 2015 hospital data that mixes ICD-9 and ICD-10);-orN/Ano data.note: the source's explanation of the flag, or a caveat such as the effect of COVID-19 on 2020 ED visits.reliability:ok,caution(a value flagged*or^^), orsuppressed(no value). Counts: 500,817 ok, 21,958 caution, 9,629 suppressed.sourceConflict: see caveats.
Finding a topic
Find the Indicator by name, follow measureIds to pick the measure you want (check measurementType and displayType), then take the Observations with that measureId. Filter by geoType to get one neighborhood system.
Heat
| Indicator | Geographies | Years |
|---|---|---|
| 2074 Heat stress: deaths | Citywide | 1999 to 2020 |
| 2075 Heat stress: yearly emergency department visits (number, rate, age-adjusted rate per 100,000) | Citywide, Borough, UHF42, CD | 2005 to 2022 |
| 2076 Heat stress: yearly hospitalizations | Citywide, Borough, UHF42, CD | 2000 to 2022 |
| 2443 Heat stress: 5-year emergency department visits | Citywide, Borough, CD | 2018-22 |
| 2410 Heat stress: 10-year hospitalizations | Citywide, Borough, CD | 2013-22 |
| 2084 Heat events (days at or above 90, 95, and 100 degrees) | Citywide | 2000 to 2018 |
| 2141 Daytime summer surface temperature | Borough, UHF42, CD, NTA2010 | 2018 only |
| 2157 Tree canopy cover | Borough, UHF42 | 2014 only |
| 2185 Household air conditioning | Borough, UHF42, Subboro, PUMA2010, PUMA2020 | 2014 to 2023 |
| 2104 Adults with air conditioners; 2155 Older adults with air conditioners | Borough, UHF34, UHF42 | 2007 to 2022; 2013 |
| 2191 Heat vulnerability index (score 1 to 5) | CDTA2020 | 2023 |
| 2411 Heat vulnerability index (NTA) | NTA2020 | 2023 |
Air quality. The neighborhood air pollution estimates from the NYC Community Air Survey: 2023 Fine particles (PM 2.5), 2025 Nitrogen dioxide (NO2), 2027 Ozone (O3), 2024 Black carbon, and 2026 Sulfur dioxide (SO2), for Citywide, Borough, UHF34, UHF42, and CD, 2009 to 2025. Annual, summer, and winter means are separate measures that share the same year time period. Also 2449 Regulatory PM2.5 monitoring and 55, 57, and 92 (ozone, PM2.5, and air quality action days), citywide only.
Health effects attributed to air pollution (modeled, multi-year): 2108 Deaths due to PM2.5, 2117 Asthma ED visits due to PM2.5, 2119 and 2120 hospitalizations due to PM2.5, 2121 deaths due to ozone, 2122 Asthma ED visits due to ozone, 2124 Asthma hospitalizations due to ozone. Citywide, Borough, and UHF42, 2005 to 2019.
Asthma
- Emergency department visits: 2048 (age 4 and under), 2379 (age 5 to 17), 2380 (adults). Citywide, Borough, UHF42, CD, 2005 to 2023.
- Hospitalizations: 84 (age 4 and under), 2381 (age 5 to 17), 2382 (adults). Same geographies, 1998 or 2005 to 2023.
- By neighborhood (NTA2010, multi-year averages): 2378, 2383, and 2384 (ED visits), 2385, 2386, and 2387 (hospitalizations).
- Survey prevalence: 18 Adults with a recent asthma attack and 2414 Adults with asthma (UHF34, UHF42); 2392, 2429, and 2431 for children (NYC KIDS survey areas).
For matching heat or air quality against health outcomes at the neighborhood level, UHF42 is the system most of these share. Tree canopy, surface temperature, heat ED visits, heat hospitalizations, the air survey, and asthma ED visits all have UHF42 values, though not for the same years.
Joining geographies to nyc/geographies
Geography ids are <GeoType>:<GeoID> with EHDP's GeoType and GeoID copied verbatim. The nyc/geographies collection uses the same id scheme for its boundaries, so Observation.geographyId and Geography ids join to it directly.
The GeoIDs are EHDP's numeric codes, not the Department of City Planning codes. NTA2020 50101 is DCP BX0101; CD 101 is Manhattan Community District 1. For NTA2010, NTA2020, and CDTA2020, the dcpCode field carries the DCP code (BX01, BX0101, BX01) from the portal's own boundary files, for joining to other city data.
boroughId uses EHDP's order: 1 Bronx, 2 Brooklyn, 3 Manhattan, 4 Queens, 5 Staten Island. DCP uses a different order (1 Manhattan, 2 Bronx, 3 Brooklyn), and CD ids follow DCP's. Use the borough name when in doubt.
Two geography types appear in the data but not in EHDP's GeoLookup, so their records have inGeoLookup: false and derived names:
UHF33(three smoking indicators, 2002 only): an older UHF grouping whose ids join the UHF42 ids they merge, e.g.UHF33:207208is UHF42 207 and 208. Names come from UHF34 where the id matches, otherwise from the UHF42 parts.National:1(two work-related death indicators): the United States.
Caveats
- Counts, rates, and age-adjusted rates are different measures. Most health indicators have a
Numbermeasure and one or more rates. Compare places with rates, and prefer age-adjusted rates when the places have different age structures. Units are indisplayType(per 100,000, per 10,000, ppb, percent of land area). - Multi-year periods. Many neighborhood estimates cover several years pooled (
TimePeriod.kindmultiYear,isMultiYeartrue), and their rates are usually average annual rates. Do not line them up with single years. The sourcetimeTypeis unreliable here: periods like2009-13and2015-2017are typedyear, so usekind. Labels are not unique:2012-16,2013-17,2014-18,2015-19,2016-20,2017-21, and2018-22each belong to two TimePeriod ids. Three periods (2010-2011,2011-2012,2012-2013, no observations) have source dates that cover a single day. - Suppressed and unstable values. Small counts are suppressed (fewer than 11 hospital or ED cases under SPARCS rules) and small or noisy estimates are flagged. Check
reliabilitybefore using a value. Suppressed observations are kept, with novalue, so you can see the gap. - Geography vintages. NTA2010 and NTA2020 are different neighborhood maps, as are PUMA2010 and PUMA2020; ids are not comparable across vintages. NYC KIDS areas change by survey year (NYCKIDS2017, 2019, 2021, 2023). Subboro areas are the older name for PUMA-based areas. Several CD-level values are shared by two community districts because the source reports them for a merged PUMA.
- Coding changes. Hospital and ED data switched from ICD-9 to ICD-10 in October 2015, and 2020 ED visits dropped with COVID-19. The
notefield and the measure'showCalculatedsay so where it applies. - Conflicting source rows. The source has 15 pairs of rows with the same measure, geography, and time period but different values (Subboro 31 in 2021 and 2023 for seven housing survey measures, and the citywide 2020 values for indicator 84, asthma hospitalizations age 4 and under). Both rows are kept: the first with the plain id, the second with
#2appended, and both carrysourceConflict: true. Neither is known to be right. - Left out. Two data files (
2188.json,2403.json, 548 rows) have no indicator in the portal metadata, so their measures have no name or definition. Some comparisons list measures or indicators that are not in the metadata; those ids are kept as plain strings inunresolvedMeasureIdsandunresolvedIndicatorIds. Portal display settings beyond which geographies are mapped and charted are not carried.