2015 LMF: Methodology & Analytic Considerations Data Release Date: November 6, 2017
Page 13 of 24 Document Version Date: April 11, 2019
middle initial (including blank), birth year (±3 years), birth day, sex, race, marital status,
and state of birth.
Type B: Records in this category were initially put in Class 5 but switched to Class 3 if, after
review, there was the possibility that either SSN was recorded incorrectly or that the
spouse’s SSN was recorded instead of the subject’s SSN. In this category, SSN was known
but three or more (of nine) and one or more (of four) digits did not agree, but at least
eight of the following items agreed: first name (including NYSIIS/Soundex match),
middle initial (including blank), last name (including NYSIIS/Soundex match), birth
year (±3 years), birth day, sex, race, marital status, and state of birth. All total scores
were adjusted to reflect the final class code for the potential matches. For example, any
record that was switched from Class 5 to Class 3 had its score adjusted to reflect that SSN
is missing, with the value of 0 assigned to SSN.
Class 4: SSN was unknown on either the NCHS survey submission record or the NDI record
and fewer than eight of the items listed in either of the Class 3 types matched.
Class 5: SSN was present but fewer than 7 (of 9) or 4 (of 4) digits on SSN agreed, and the record
did not fall into a prior class.
4.6 Selecting Matches and Assigning Vital Status
Since each eligible NCHS survey participant may have had multiple submission records and
each submission record may have returned one or more potential matches to a NDI record,
NCHS SPB employed a strategy to provide the single best NDI match record for inclusion on the
linked mortality file. First, NCHS-NDI potential match records that had a date of death prior to
the date of interview or a score less than or equal to zero were considered false matches and
were eliminated from the pool of potential matches. Many participants, however, still had more
than one NDI record as a potential match, and different records could potentially end up in
different classes. The remaining potential matches were ranked first on class (from 1 to 4) and
then within class by highest score (Note: Class 5 records were determined non-matches). NCHS
SPB selected the NDI match with the highest score within the best class (if in class 1 or 2) or the
highest score only (if in class 3 or 4). In the event of a tie among NDI record matches for a
particular NCHS survey record, the matching criteria were compared and the record with the
most matching criteria was selected.
Next, NCHS SPB determined whether each best record was a match. A match reflects both the
best match for vital status of the survey participant and a corresponding match to the correct
death certificate data. All class 1 match records were considered matches. Within each class,
matches with a score greater than or equal to the cut-off score were considered matches whereas
records with a score less than the cut-off were considered non-matches. The cut-off scores for
classes 2, 3, and 4 were 44, 45, and 42, respectively. The cut-off scores within classes 2, 3, and 4
simultaneously maximized the proportion of people correctly classified and minimized the
number of people incorrectly classified, with particular attention given to minimizing the
number of false positives.