Infant Mortality by County

×

What you're seeing. Infant mortality is deaths before a first birthday per 1,000 live births. The numbers come straight from CDC WONDER, not from a dashboard or an aggregator, and every rate is recomputed from the underlying counts rather than copied from WONDER's printed rate. Rates are pooled over a five-year window because a single year of births in most counties is too few to say anything.

Lower is better here, which is the opposite of the LE Gap by Age page. Green means a county buries fewer infants per 1,000 births than the country does; amber means more. Deeper colour is a bigger difference; anything within about a seventh of a point of the U.S. stays gray.

Two methods, and the Tier column says which. Tier 1 is the NCHS pre-linked file (WONDER D69), where each infant death is matched to its own birth certificate — exact, but NCHS will not identify a county under 250,000 people, so it reaches only 261 of them. Tier 2 fills the gap by linking the separate NCHS mortality and natality files at county level instead: mortality-file under-1 deaths over live births. It runs about 4% off tier 1 where both exist — the price of linking at the aggregate rather than at the record — and it roughly doubles coverage. Rank within a tier, not across them.

The counties you cannot see are worse than the ones you can. This is the caveat that has to travel with every number on this page. The named counties are the country's metropolitan core, not a random sample of it: in the current window they run 5.20 against 6.12 for everywhere else, and that gap has widened from +9.1% to +17.9% since 2007-2011. A map or a trend built from these counties alone reads better than the country actually is, and increasingly so.

Why there are two deaths columns. No single source of infant deaths covers every county here: the linked file names only the 261 counties NCHS identifies, and the mortality files are what tier 2 counts instead. So each tier gets its own column, filled only on its own rows — 261 + 306 = 567, no overlap and no gaps. Every county carries exactly one death count, and the column it sits in is the method, so there is nothing to cross-reference. Take that count, divide by births, multiply by 1,000, and you get the rate printed beside it — the counts and the rate are the same arithmetic.

The two columns are not a series. Where both methods can be run they land a median 2.1% apart, up to 10.9% — different linkage, different weighting. Neither is a correction of the other, and a tier-1 count and a tier-2 count are not comparable quantities, so do not difference them, sum them, or rank across them. Every number in either column is an observed count; no substituted state rate reaches this page.

Watch the time base. Population is an annual average; births and deaths are five-year totals, because that is the scale the rate is built on. The labels say so, and the mismatch is deliberate — the population file stores person-years (48.9 million for Los Angeles, not 9.8), and a column reading 48.9M next to a rate would mislead worse.

Two flags worth respecting. The Flag column marks a rate resting on fewer than 20 deaths — too few to be stable — read that rate next to this column or not at all. It is computed from each row's own death count, so it means the same thing on both tiers, and WONDER's own marker is honoured as well and never overridden. And vs prior compares two windows that share four of their five years, so it is one new data year moving an average, not an independent period-over-period change. Blank means the county carries a rate in only one of the two windows: a roster change, not an improvement.

Not the same as “Infant Mortality (Pop)”. Elsewhere on this site a second infant column divides infant deaths by under-1 resident population instead of live births. Nationally the two land four hundredths apart, which is a coincidence of aggregation and not corroboration — at county level they diverge by a median 5.4%, and the divergence tracks how wrong the imputed population denominator is. Do not merge them, do not average them, and never shorten either to “infant rate”.

If your county isn't here. Around 2,600 counties carry no rate on this page, because neither method can reach them — NCHS will not identify a county under 250,000 in the linked file, and the county-linked method needs an observed death count it does not always have. For those, HRSA's Maternal and Infant Health Mapping Tool maps infant mortality — and low birth weight and other birth outcomes — for nearly every county. It gets there by spatial smoothing, not by measuring more, so a county value there is a modelled estimate borrowing from its neighbours rather than a count, and it will not match the measured rates on this page. Two windows only, 2017-2019 and 2020-2022.

Source: CDC WONDER Linked Birth / Infant Death Records (D69) for tier 1; CDC WONDER mortality files (D77/D157) over Natality (D66) births for tier 2. Population from the life-expectancy file, shown as an annual average. Compiled on DemographyInfo.org.