CDC Data.CDC.gov Open Data Portal

Datadory ships the CDC's entire public-health catalog as one maintained feed - about 1,078 dataset resources, roughly 225 of them tagged to the National Center for Health Statistics, spanning every program office's output rather than one agency corner. Six families carry the analytical weight:

• Vital statistics - births, deaths, life expectancy at birth, teenage birth rates, general fertility, with quarterly provisional estimates alongside settled annual history.
• Mortality detail - leading causes of death under the NCHS 113 grouping, age-adjusted death rates by state, drug-poisoning mortality mapped at county level.
• Chronic disease - the indicator system tracking prevalence of diabetes, heart disease, cancer screening and their risk factors across states and counties.
• Immunization - coverage estimates by vaccine series and population group.
• Surveillance - infectious-disease monitoring and behavioral risk factor series.
• Social determinants - insurance coverage, access delays and the socioeconomic context that moves everything above.

For health care REIT work this is the demand side in federal grade: aging, chronic-disease burden and county mortality differentials are precisely the inputs senior housing, medical outpatient and post-acute absorption models consume.

Get a sample of this dataset - name the topics, geographies and span, and we return real rows cut from these files.

Two records, two rhythms - shown flat, exactly as they arrive:

# Quarterly provisional birth indicators -- 2026 Q1
year_and_quarter   2026 Q1
topic              Birth Rates
topic_subgroup     Teen Birth Rates
indicator          10-14 years
race_ethnicity     All races and origins
rate               0.1
unit               per 1,000 population
significant        *

# Leading causes of death -- settled final-year file
year               2012
state              Vermont
_113_cause_name    Nephritis, nephrotic syndrome and nephrosis
                   (N00-N07,N17-N19,N25-N27)
cause_name         Kidney disease
deaths             21
aadr               2.6

Read together they explain why the catalog earns shelf space. The first row is provisional surveillance: a 2026 Q1 estimate published with a significance marker while the numbers settle. The second is settled fact - twenty-one Vermont kidney-disease deaths in 2012 at an age-adjusted 2.6 per 100,000, safe to build a decade-spanning cohort study on today. Very few collections hold both ends of that spectrum under one field vocabulary.

Thirteen documented fields cover the two record families sampled above, each verified against its published column definition. On the mortality side _113_cause_name is the load-bearing column: because NCHS keeps the 113-causes grouping comparable across decades, a state panel built on it survives vintage changes without remapping. On the quarterly side topic and topic_subgroup do the same job for birth indicators, race_ethnicity carries the stratification that turns a national rate into a demographic story, and significant flags which provisional estimates are still moving.

The complete dictionary follows. Program areas outside natality and mortality publish their own column sets, so those fold into the request step rather than cluttering every pull - see additional fields on request beneath the table.

Three chips, honestly drawn:

• Geography - United States end to end. National totals and all states throughout; county detail on selected panels such as drug-poisoning mortality; some sub-county geographies on specific resources.
• Temporal - the deep end is what surprises people. Mortality-rate series begin in 1915, annual vital statistics run continuously since, and quarterly provisional estimates were observed through 2026 Q1 at research time. A handful of surveillance series end at their final release rather than extending forward, which we surface per resource rather than papering over.
• Granularity - aggregate counts and rates keyed by geography x period x demographic stratum. No record-level microdata anywhere: these are designed-for-publication tables, which is exactly what a demand model wants and what privacy review approves fastest.

Set against the wider Datadory catalog - 1,744 datasets across 159 industries averaging 7.81 - this record scores 9/10, carried by verified field documentation and a temporal run few commercial vendors attempt.

API, files, or your warehouse. Daily, weekly, or hourly.

You pick the channel and the cadence; the field dictionary above travels unchanged through all three. Flat files suit overnight warehouse loads and BI tools, structured responses suit dashboards that refresh themselves, and warehouse delivery puts federal vital statistics one join away from your property, membership or claims tables. Changing cadence is a settings conversation, not a re-integration project - and a sample cut to your topics and geography comes first either way.

• Senior-housing demand sizing. Life expectancy at birth plus the chronic-disease and mortality mix by state quantify how long and how healthily the 75-plus cohort lives - the numerator of every independent-living and assisted-living absorption model.
• Medical outpatient site selection. Access-delay and chronic-disease indicators flag where routine care already strains, which is where new medical office space tends to lease first.
• Epidemiological benchmarking. Age-adjusted death rates under the NCHS 113 grouping give actuaries and underwriters a federal yardstick for portfolio morbidity assumptions.
• County-risk screening. Drug-poisoning mortality mapped to county surfaces community-health stress that raw population counts hide, sharpening acquisition and disposition screens.
• Thesis-grade macro overlays. A century of demographic-health history lets investors test whether a demand assumption held through the last four cycles before betting the next one.

Investors and quants hold it at relevance 2 - the 1915-forward run gives REIT demand theses evidence that outlives any earnings cycle. Data scientists also score 2: aggregate tables clean enough to join straight onto property or claims data without de-identification work. Market researchers and consultants get citable federal numbers for sizing decks. Journalists and academics get the standard citation for health-services research at whatever cadence a newsroom or longitudinal study needs. Developers building health-adjacent products and competitive-intel teams round out the tag list - the former embedding indicators on a stable schema, the latter reading new surveillance topics as demand signals.

• CDC NCHS Data Browsing Collection - the curated ~225-resource NCHS slice of this same catalog. When vital statistics is the whole job, the smaller collection moves less noise; when breadth matters, this portal wins.
• CMS Data Hub - Medicare & Medicaid Datasets - the payer side of the same coin: enrollment, spending and utilization where this catalog supplies health status. The head-to-head scores both field by field.
• Dartmouth Atlas of Health Care - regional Medicare utilization across about 306 hospital referral regions, a small-area utilization lens this catalog does not attempt.

Browse the ranking on best health care reits datasets, or start from the health care reits data hub.

Three things, stated up front.

First, this is aggregate data by design: counts and rates keyed to geography and demographic stratum, never individual records, and county-level cuts exist on selected panels rather than across the board. If your model needs encounter-level detail, pair this with an encounter source rather than expecting it here.

Second, provisional and settled coexist. Quarterly estimates carry significance flags until they finalize; annual and historical files are settled fact. Treat the newest quarters as directional and the long series as bedrock - the sample shows you which is which for your chosen topics.

Third, breadth varies by program area. The natality and mortality families share the thirteen-field vocabulary above; other areas publish their own column sets, so tell us the topics and span you want and the sample gets cut from the right resources - with the exact column names your pipeline will key on - rather than the first match.