14 National Vital Statistics Reports, Vol. 69, No. 2, January 30, 2020
was to see if the records would have some indication of a
pregnancy-related hospital encounter, including delivery, or that
a pregnancy would be noted in their record of care for a given
inpatient or emergency department visit. Pregnancy-related
hospital encounters were defined based on any diagnosis codes
using International Classification of Diseases, Ninth Revision or
10th Revision, Clinical Modification (ICD–9–CM or ICD–10–CM)
or SNOMED for that encounter. If no indication of a pregnancy-
related diagnosis in the previous year was listed for these
decedents, then these records would be considered a potential
overuse of the pregnancy indication on the checkbox, or a
false positive use of the checkbox (i.e., the checkbox indicated
pregnancy when it should not have).
Underuse of the pregnancy indication on the checkbox
was assessed by first identifying all records of pregnancy-
related hospitalizations for persons who died within the year of
their hospitalization and then determining those records with
a pregnancy checkbox value of 1, 7, 8, or 9. Given that these
decedents had a record of a pregnancy-related hospitalization
in the previous year, the expectation is that the checkbox value
would indicate a current or recent pregnancy. Otherwise, if
no value indicating current or recent pregnancy existed in the
checkbox item, these records were considered an underuse of
the pregnancy indication of the checkbox, or a false negative
use of the checkbox (i.e., the checkbox should have indicated a
pregnancy when it did not).
The 2014 NHCS data were linked to the 2014–2015 NDI, and
2016 NHCS data were linked to the 2016–2017 NDI, to allow for an
assessment of mortality at intervals of interest (e.g., death at 30,
60, or 90 days postdischarge or 1-year survival postdischarge).
The linkage process has been detailed elsewhere: https://
www.cdc.gov/nchs/data/datalinkage/NHCS14_NDI14_15_
Methodology_Analytic_Consider.pdf and https://www.cdc.
gov/nchs/data/datalinkage/NHCS16_NDI16_17_Methodology_
Analytic_Consider.pdf.
Note that NHCS data for both 2014 and 2016 are based only
on the relatively small percentage of hospitals that responded to
the survey (16% in 2014 and 27% in 2016) and are not nationally
representative. Therefore, this case study is not intended to be
a nationally representative assessment of checkbox usage.
In addition, this analysis only includes visits that occurred at
sampled hospitals, and the possibility exists that the patient may
have been seen primarily at a nonsampled hospital during the year
before birth or that the mention of a pregnancy was omitted from
the record. Lastly, this analysis may miss cases of pregnancies
for which no event occurred necessitating a hospitalization or
emergency department visit (e.g., abortive outcomes).
Potential overuse of checkbox—false positives
To create the analytic sample to assess overuse of the
pregnancy checkbox, 150 women in the 2014 linked NHCS–NDI
and 220 women in the 2016 linked NHCS–NDI file were identified
with a checkbox value of 2, 3, or 4. In the 2014 NHCS, 81 of
the 150 decedents who were identified as pregnant within the
year before death by the checkbox (values 2, 3, or 4) had no
pregnancy-related hospital encounter within 1 year of death,
including no record of delivery. Of these 81 decedents, 32 were
under age 30, 15 were aged 30–39, and 34 were aged 40 and
over. Based on this analysis, 54% of records with a pregnancy
checkbox value indicating a current or recent pregnancy had no
pregnancy-related hospitalization nor any mention of pregnancy
in the hospitalization record in the previous year and were
considered false positives.
In the 2016 linked NHCS–NDI data, 124 of the 220 decedents
who were identified as pregnant within the year before death by
the pregnancy checkbox (values 2, 3, or 4) had no pregnancy-
related hospital encounter within 1 year of death. Of the 124
decedents, 47 were aged under 30, 36 were aged 30–39, and 41
were aged 40 and over. Based on this analysis, 56% of records
with a pregnancy checkbox value indicating a current or recent
pregnancy had no pregnancy-related hospitalization nor mention
of pregnancy in the record of hospitalization in the previous year
and were considered false positives.
Potential underuse of checkbox—false negatives
In the 2014 NHCS, 236 decedents with a pregnancy-related
hospitalization in the previous year were identified. Of these,
157 decedents had both a pregnancy-related hospital encounter
within 1 year of death and a pregnancy checkbox value of 1, 7, 8,
or 9 (not indicating pregnancy). Of the 157 decedents, 62 were
aged under 30, 62 were aged 30–39, and 33 were aged 40 and
over. Based on this analysis, 67% of 236 records of a pregnancy-
related hospitalization in the previous year did not have current
or recent pregnancy indicated by the pregnancy checkbox on the
death certificate and were considered false negatives.
In the 2016 NHCS, 253 decedents with a pregnancy-related
hospitalization in the previous year were identified. Of these,
158 had both a pregnancy-related encounter within 1 year of
death and a pregnancy checkbox value of 1, 7, 8, or 9. Of the
158 decedents, 64 were under age 30, 62 were aged 30–39, and
32 were aged 40 and over. Based on this analysis, 62% of 236
records with a pregnancy-related hospitalization in the previous
year did not have current or recent pregnancy indicated by the
pregnancy checkbox and were considered false negatives.
Summary
This analysis attempts to discern use of the pregnancy
checkbox from a data source linking hospitalization encounters
and deaths among patients. The purpose was to assess the use
of the checkbox according to whether a record of pregnancy in a
hospitalization event occurred in the period of time defined by the
checkbox (i.e., within 1 year). For both the 2014 and 2016 linked
files, approximately 53% of records were false positive cases of
pregnancy indication on the checkbox, while approximately 65%
of records were false negative cases of pregnancy indication
on the checkbox. The ability to examine use of the pregnancy
checkbox on the death certificate linked with hospitalization data
predating death provided a unique way to examine and highlight
potential misreporting of the checkbox, which may, in turn, affect
estimates of maternal mortality.