Why infant mortality figures arrive three months late

The National Center for Health Statistics publishes **quarterly provisional
estimates of infant mortality**, built from birth and death certificate data
received and processed as of a stated cutoff date.

They are provisional in a specific sense, and the reasons are worth
understanding before reading any single quarter's number.

Why the lag

**Approximately one-third of infant death records are initially received with
the cause of death undetermined, or pending investigation.** Infant deaths are
often subject to lengthy death scene investigations and autopsy, including
ancillary tests such as toxicology and genetic testing.

So estimates are provided with a three-month lag from the close of the
quarter, to allow those investigations to finish. Publishing sooner would mean
publishing a third of the deaths as "cause unknown".

Estimates are also subject to revision in future quarterly releases as more
records for that quarter arrive, and individual records may be weighted to
independent provisional counts of all infant deaths in each state by month, to
adjust for incompleteness.

What is measured

Plus death rates for the five leading causes of infant death: congenital
malformations, deformations and chromosomal abnormalities; disorders related
to short gestation and low birth weight; newborn affected by maternal
complications of pregnancy; sudden infant death syndrome; and unintentional
injuries.

Those five were chosen for their relevance to surveillance and because
reliable estimates can be produced from provisional data — other causes may be
added in future releases.

The rate you are reading is a moving average

Rates are presented for the 12-month period ending with each quarter — a
12-month moving average — to account for seasonality. **Quarterly three-month
estimates are not provided at all**, because they would be unstable and
seasonal swings in births and infant deaths would dominate them.

Each 12-month figure is compared with the same period a year earlier, and
differences are reported only if significant (p ≤ 0.05).

That is the discipline behind the headline: a change you see reported here has
survived a seasonality adjustment and a significance test, which is not true
of most quarter-on-quarter statistics.

*Source: National Center for Health Statistics, Centers for Disease Control
and Prevention.*