Hi IPUMS, I’m conducting an analysis using the pooled 1986-2014 NHIS-NDI linked mortality files. I’d like to calculate follow-up time for respondents in the analysis. Quarter and year of death are provided by ‘MORTDODQ and ‘MORTDODY’, respectively. To identify quarter and year of the baseline interview (needed to calculate time on study), it looks I like should use ‘INTERVWMO’ and ‘INTERVWYR’ from 1997-2018. Before 1997, when INTERVWYR is unavailable, should I assume the baseline interview occurred in the number provided in the variable ‘YEAR’? I noticed that INTERVWYR and YEAR don’t always align, so relying on YEAR throughout doesn’t seem like a perfect option. Thanks for the help!
I consulted with the IPUMS NHIS team on this, and they advise that you employ YEAR before 1997 and INTERVWYR in 1997 and later years when working with dates in the NHIS. INTERVWYR only became available in the original data starting in 1997. The NCHS documentation between 1997-2003 notes that a few interviews were completed during January of the following year, so in those cases the interview year will differ from the sample year. Before 1997, the sample year is the only year date information available. They have also been unable to find any mention in the original documentation where there is a reason given for why NHIS started to offer the INTERVWYR variable on the public use file in 1997 but not before.