USER GUIDE

Community Health Assessment User Guide

A multi-project analytical platform for community health assessment, built around real reference cases across the country.

What this is

The Community Health Assessment tool turns Census-tract-level public data into a Community Health Needs Assessment (CHNA) evidence base for any hospital, county, or AIAN reservation anchoring the analysis: a defined community, a benchmarked health profile, a tract-level disparity explorer, a transparent candidate significant-needs ranking, an AI-written multi-factor analysis tool, and a build-your-own chart tool.

It is the analytical core of a CHNA workflow. It does not yet include community engagement, prioritization workshops, resource inventory, implementation strategy, or impact monitoring. Those are real future capabilities, deliberately deferred rather than half-built. See Scope and limitations below.

Projects

Everything in the tool lives inside a project: one state (or set of states), an anchor (a hospital, a county, or a reservation), and a chosen set of factor domains or measures. A project is built once and reopens instantly afterward, rather than being rebuilt on every visit.

The Projects landing page lists every project with its status, anchor, and states, and is where a new one gets built. New Project starts with a Project Type choice:

Hospital: pick the state or states, then search HIFLD's open-hospital registry (which also includes IHS and Tribal facilities). Optionally check for a Neighborhood map.
County: search and select one or more counties nationwide by name, with no hospital involved, even across state lines (for example, for a regional comparison). Selecting exactly one county gives the project a real home county for benchmarking (full County, State, and National, the same as a hospital project). Selecting more than one falls back to State and National only, since there is no principled way to pick one as home.
Reservation: search and select one or more federally or state-recognized AIAN reservations or tribal statistical areas directly, with no hospital involved. The states needed are derived automatically from the reservation's own footprint; selecting Navajo Nation, for example, pulls in Arizona, New Mexico, and Utah without you needing to know that in advance. Select more than one entry to include a tribe's full extent (a reservation and its separate off-reservation trust land each have their own entry) or to combine several tribes in one project.

Then choose which factor domains (or individual measures) to include; Select All and Deselect All shortcut the checklist at either level. Click Build. A background build takes roughly 1 to 10 minutes depending on state size. A progress page updates until it is done and then opens the new project's Overview.

Edit (from a project's card) changes an already-built project's factor selection, and for a hospital-anchored project its default hospital, without a full rebuild: seconds to about a minute, not the original build time. The same Select All and Deselect All shortcuts apply. A county- or reservation-anchored project's anchor is fixed at build time.
Switch between built projects at any time from the Project selector in the top navigation.

Navigation

Pages in the Community Health Assessment tool
Page	What it shows
Overview	The defined community (Census tracts, population, method, and sources) on a map, with headline measures and a top candidate need. The map follows the topic picked in the Community Narrative Summary dropdown: Population for Demographic Overview, or that domain's composite need for any other selection. It always overlays every open hospital and health facility (HIFLD and IHS ITU) in the project's states as points, whatever the anchor.
Community Profile	Every curated factor's community value against county, state, and national benchmarks, grouped by domain, with a trend arrow (and a trend line on hover) for each factor. A Top 20 Highest-Importance Measures table at the top ranks individual measures by Evidence Score, ahead of the full domain-by-domain tables.
Disparity Explorer	A tract-level map for any curated factor, with a multi-year trend chart, high-need tract counts, population affected, and the highest-need reliable tract. Projects that touch a reservation also get a Reservations map view, showing every reservation in the project's states side by side.
Significant Needs	Domain-level candidate needs ranked by a transparent Evidence Score, with every component factor shown. Never a black-box composite.
Tract Explorer	A sortable table with one row per community tract (or rolled up by town or city), with a Composite Need Score and a column per domain. Each number is colored on the same blue-to-red decile scale as the maps, so better and worse values read at a glance. Extra views: Neighborhood for projects with a neighborhood layer, Reservation Comparison for projects that touch a reservation, and County Comparison for county-anchored projects, each showing every neighborhood, reservation, or county in the project's states side by side.
Graph Builder	Build and save custom charts (line, column, bar, scatter, dumbbell, and box-and-whisker) for any curated measure, with flexible breakdown, axis, and color options. Reopen a saved chart at any time from this page.
Analysis	Pick one or more curated factors and, optionally, describe what the write-up should focus on (for example, "which tracts should we prioritize first?") for an AI-written deep dive grounded in the community's own profile and disparity data. Leave the focus box blank for a standard benchmark-comparison and disparity write-up. Exportable as TXT, and saved and reopened like every other saved item.
Library	Every saved item in the project (analysis write-ups, chats, summaries, graphs, maps, and CSV exports from Community Profile and Tract Explorer) plus uploaded local files (surveys, meeting notes, letters), grouped by type with a checkbox on each. See Library and Build Report below.

An AI assistant is available from a floating button on every page. It answers questions grounded in the dashboard's own data by default, falls back to a web search for anything outside that scope, and labels which source it used. Within a project, a conversation can be named and saved, then reopened or deleted from a Saved list without leaving the chat panel.

Library and Build Report

The Library page is where every saved item in a project ends up, so you can reopen, curate, or delete from one place instead of hunting across pages.

Send to Report Builder flags checked items as grounding material for the next report. Remove from Report Builder clears the flag without deleting anything.
Delete removes checked items outright.
Upload File adds a local document (a survey export, a scanned letter, meeting notes) as is. The tool never parses or displays its content.
Download Grounding Bundle packages the project's own verified figures (factor table, prioritized needs, and available resources) plus every flagged item into one ZIP file, for use outside the tool.

Build Report runs an AI agent in the background, chapter by chapter, against every flagged item plus the project's own verified factor data, so nothing needs to be re-checked when you click. It runs as a tracked background job with a live progress bar. A finished run appears in the Reports section of the Library as a downloadable ZIP of DOCX chapters plus a compiled HTML report. Runtime grows with the number of chapters and the size of the project; a large multi-county project can take a couple of hours.

Defining your community: hospital, county, and reservation anchors

For a hospital-anchored project, every page shares one control bar at the top. Changing it recomputes the whole dashboard live; nothing needs to be rebuilt.

Hospital: a searchable list of every open hospital HIFLD has on file for the project's states, including IHS and Tribal facilities.
Area of Operations: how the hospital's community is defined. There are five methods, switchable on any page.

Area of Operations methods
Method	How it defines the community
County	One or more counties, chosen from a checklist. It starts with the hospital's home county; check more to expand the community, for example to cover a documented multi-county service area.
MSA	One or more Metropolitan Statistical Areas, with the same checklist. It starts with the hospital's home MSA.
Radius	Every tract whose center falls within a chosen radius, in miles, of the hospital.
Neighborhood	Offered only for projects with a Census-tract-aligned neighborhood layer (see the Boston example below). Pick one or more named neighborhoods directly.
Reservation	Offered only when the project includes at least one Census tract with a majority (over 50 percent) land-area overlap with an AIAN reservation or tribal statistical area. Pick one or more directly. This lets a hospital scope its community to a reservation it serves without a separate reservation-anchored project.

A checkmark icon and a count on the County, MSA, Neighborhood, and Reservation buttons (for example, "Suffolk · 1/14") means a full checklist sits behind the button, not one fixed value. Click it to add or remove areas.
The Census Tract and Town/City toggle (on Overview, Disparity Explorer, Significant Needs, and Tract Explorer) switches every map and table between tract geometry and a population-weighted rollup to municipal boundaries.

A county- or reservation-anchored project shows a simpler control bar. There is no hospital to pick and no Area of Operations to switch, because the counties or reservations chosen at build time are the community. It keeps one control: an Included Counties or Included Reservations checklist, listing every county or reservation in the project's states (not only those chosen at build time). It starts with the build-time selection and can be widened or narrowed live, recomputing without a rebuild, just like a hospital project's County and MSA checklists. The checklist appears only when the project's states contain more than one county or reservation to choose from.

For both of these anchors, Tract Explorer's Distance column stays empty, since there is no single hospital location to measure from. In Community Profile, a reservation-anchored project always compares against State and National only (no County column, since a reservation can span counties or states). A county-anchored project gets a County column too, but only while exactly one county is included; with more than one, it uses State and National only.

Methodology

Aggregation: community, county, state, and national values are population-weighted averages (for rates and indices) or sums (for totals), never a simple average of tract percentages. Missing tract values are left out of the weighted average, not treated as zero.
Evidence Score: for each score-eligible factor, Evidence Score = Burden + Benchmark Gap + Disparity.
Burden is the community's need percentile among all of the project's tracts for that factor.
Benchmark Gap is the need-oriented percent difference from the national value, floored at zero, so doing better than the national benchmark never counts against a factor.
Disparity is the spread between the 90th and 10th percentiles of tract need within the community.
Domain-level scores are the average of their component factors, and every Need Card shows the components behind it.

Indicator roles: every factor is tagged Outcome, Behavior, Prevention, Access, SDOH Need, Environment, Disability, Context, or Stratifier. Context and Stratifier factors (demographic composition such as race, sex, and ethnicity, which can describe who is affected by a disparity) never score. Every score-eligible factor falls under one of the other seven roles, but the role alone does not make a factor eligible: eligibility is a separate curation decision for each factor. Only 99 of the catalog's 432 factors are score-eligible; see Data sources below.

Reliability guardrail: tracts with fewer than 500 residents (typically non-residential, such as airports, harbor islands, and institutional grounds) are excluded from headline "highest-need tract" picks, because modeled estimates there are statistically unstable. They stay visible on the map.

Data sources

Data sources behind the Community Health Assessment tool
Source	Coverage
AHRQ Community-Level Health Database (ACS-derived)	The bulk of the catalog (312 factors): demographic, economic, housing, insurance, and household-composition measures across 34 domains
CDC PLACES (public API)	40 factors: health outcomes, prevention and screening rates, health behaviors, and six disability types (hearing, vision, cognitive, mobility, self-care, and independent living)
CDC/ATSDR Social Vulnerability Index	21 factors: the Social Vulnerability domain and housing-cost burden
HRSA Health Professional Shortage Areas and Medically Underserved Areas	12 factors: the Underserved Areas domain
USDA SNAP-Authorized Retailer Access Map (SRAM)	11 factors: Food Access low-access tract flags and population shares
EPA Air Quality System, Emissions Inventory, and Air Toxics Screen	11 factors: Environmental Releases and Environmental Health (ozone, particulate matter, and air-toxics cancer risk)
HRSA Health Center Program sites, plus CMS, HRSA, and NPPES facility directories	12 factors: Access to Care facility counts and distance to the nearest ED-, OB-, ICU-, or trauma-capable hospital, HRSA health center, and urgent-care site
NOAA nClimGrid-Daily	8 factors: county-level Physical Environment climate measures
CDC (VSRR overdose mortality and drinking-water arsenic)	3 factors, including the county-level drug overdose mortality rate, applied to every tract in the county
AHRQ Community-Level Health Database (county-level mortality)	The county-level assault and homicide mortality rate, applied to every tract in the county
Census (2020 county population density)	1 factor
HIFLD Open Hospitals and IHS ITU Health Facilities	The hospital registry every project's Hospital selector searches
Census tract-to-AIANNH relationship file (via Oahe's shared geography assets)	The reservation and tribal-area crosswalk behind the Reservation Area of Operations method and reservation-anchored projects
Census TIGER/Line, plus Census Place and County Subdivision boundaries (via Oahe's shared geography assets)	Tract geometry and the municipal boundaries behind the Town/City toggle

432 factors across 44 domains were curated for CHNA relevance (99 score-eligible), sized to what the data supports well rather than to a fixed target. Any one project's factor selection (made at build time, and refinable with Edit) can cover fewer than the full catalog.

Boston example: a neighborhood-level community

Boston Medical Center is a founding participant in the 2025 joint Boston Community Health Needs Assessment, which defines its community as the City of Boston. The Boston project reconstructs that boundary from Census-tract-aligned neighborhood geography (Roxbury, Dorchester, Mattapan, South Boston, and others) rather than a single Boston municipal boundary. That is also what makes the Neighborhood Area of Operations method available for this project. Every new project starts with a municipal (Census Place) community for its home hospital, refinable with County, MSA, or Radius from there.

Reservation example: an anchor with no hospital

The Pine Ridge Reservation project is built with Project Type: Reservation, selecting the Pine Ridge Reservation entry with no hospital at all. The build form derived South Dakota automatically from the reservation's own geography instead of asking which state to include. The resulting community is 6 Census tracts across 3 South Dakota counties (18,847 residents).

Every analytical page (Overview, Community Profile, Disparity Explorer, Significant Needs, Tract Explorer, Graph Builder, Analysis, Library, and the narrative summaries) runs against it exactly as it would for a hospital-anchored project, with the exceptions noted above: State and National benchmarks only, and no Distance column. South Dakota has 8 federally recognized reservations and tribal areas in total. Tract Explorer's Reservation Comparison view shows all 8 side by side for context, and the control bar's Included Reservations checklist ("Pine Ridge Reservation · 1/8") can widen the live view to any of the other 7 without rebuilding anything.

Scope and limitations

Not yet built: a Community Engagement Workspace, a Prioritization Workshop, a Community Resource Inventory, an Implementation Strategy Builder, Impact Monitoring, and evaluation of actions taken since a prior CHNA.

Known caveats

Population-weighted averages of tract-level medians (for example, median household income) are a standard approximation, not a true recomputed population median.
Some social determinants measures used for national comparability (for example, households without a vehicle) read differently in a dense, transit-rich city than they do nationally. The Evidence Score surfaces this as a candidate finding, not an assertion of hardship.
The two county-level mortality measures (overdose, and assault and homicide) are applied identically to every tract in a county. They describe the county, not variation within it.
A community definition reconstructed from a hospital's publicly stated CHNA scope (like Boston's) is not confidential patient-origin data the hospital has not published.
A reservation-anchored project, or a hospital-anchored one using the Reservation Area of Operations method, has no single home county to benchmark against, since a reservation can span several counties or states. Community Profile compares these to State and National benchmarks only. A county-anchored project built on more than one county works the same way, for the same reason.

Back to Community Health Assessment