The sampling frames for the
live-birth and infant-death components
of the NMIHS consisted of the files of
birth certificates and certificates of
infant death for all registration areas
except Montana and South Dakota.
Montana did not participate in the
survey because State clearance was not
received in time to sample certificates.
After providing a limited number of
certificates for inclusion in the survey,
South Dakota withdrew clearance
because of a State law that required the
attending physician’s approval for
release of the medical record. The
sampling frame for the fetal-death
component of the NMIHS consisted of
the file of reports of late fetal death for
all registration areas except Montana,
Michigan, New York State, and South
Dakota. State laws in Michigan and
New York State prohibit the provision
of reports of fetal death for followback
surveys; therefore, those States were
eliminated from the sampling frame for
the fetal-death component of the
NMIHS. The three sampling frames
included no records from Puerto Rico,
the Virgin Islands, or Guam.
Certificates of live birth, reports of
late fetal death, and certificates of infant
death to nonresidents of the United
States were excluded from their
respective sampling frames. Attempts
were made to exclude mothers under 15
years of age from the sampling frames;
however, maternal age is not reported
on certificates of infant death, and
corresponding birth certificates were not
available until after the sample was
drawn. A small percent of the
infant-death (0.5 percent) sampling
frame included women under 15 years
of age. To decrease the possibility of
overlap of the NMIHS with CDC’s
Pregnancy Risk Assessment Monitoring
System (PRAMS), approximately
one-half of the certificates of live birth
and infant death were excluded from the
sampling frames for Maine, Michigan,
and West Virginia, which participated in
PRAMS. The District of Columbia,
Indiana, and Oklahoma sent all sample
certificates to NCHS rather than
one-half of the certificates as instructed;
weights were adjusted to account for
this sampling inconsistency. In the case
of Missouri, the live-birth and
infant-death sampling frames were
restricted to January 1 through June 30,
1988.
It is estimated that more than
99 percent of all live births and deaths
to infants that occurred in the United
States in 1988 were registered (7,8). The
number of fetal deaths that are not
registered is unknown (8), however,
underregistration is thought to increase
as gestational age decreases. The
NMIHS was designed to restrict fetal
deaths to 28 weeks gestation or more to
minimize the effect of underregistration.
The reporting of fetal deaths varies
by State or registration area, with some
reporting at all periods of gestation,
some 20 weeks gestation or more, and
some a combination of gestational age
and/or birthweight (8). Prior to selecting
the sample, approximately three in five
reports of fetal death with particular file
numbers assigned by the State or
registration area were requested by
NCHS for screening. A total of 23,483
records were screened for gestations of
28 completed weeks or more, based on
last menstrual period or physician’s
estimate of gestation. There were
instances where information on the final
fetal-death certificate differed from the
preliminary fetal-death certificate used
in screening. Of the 4,773 fetal deaths
sampled, 92 had gestational ages less
than 28 completed weeks, and 601 were
missing gestational age. These fetal
deaths should have been ineligible for
the NMIHS but were included in the
study. Gestational age was missing from
a total of 5,679 (19.3 percent) reports of
all fetal deaths in 1988 (8).
Within registration areas, sampling
strata were formed on the basis of
child’s or fetus’s race (black or other
than black), and in the live-birth
component, birthweight (less than 1,500
sampling subdomains in the live-birth
component and the two sampling
subdomains in the fetal-death and
infant-death components of the NMIHS.
Implicit stratification was employed to
ensure that samples were representative
with respect to maternal and infant
characteristics. Within each sampling
stratum in the live-birth and fetal-death
components, records were sorted by
mother’s marital status and age before
being selected for the sample. In the
infant-death component, records were
sorted by child’s sex and age at death
within each sampling stratum.
According to the vital registration
system for 1988, 4,234 certificates of
live birth lacked a response for
birthweight (7). In addition, data items
used to form poststratification cells may
have been missing because vital records
used to draw the NMIHS sample were
not imputed. The following conventions
were used when a data item used to