Using the National Death Index to obtain underlying cause of death codes

This study evaluated the comparability of underlying cause of death codes obtained from NDI Plus, a new feature of the National Death Index (NDI), with codes assigned by two study nosologists or by a National Center for Health Statistics (NCHS) nosologist. Two study nosologists and an NCHS nosologist independently reviewed the death certificates of 493 decedents and assigned each an International Classification of Diseases code for the underlying cause of death. Using the NCHS codes as the reference standard, we determined discrepancy rates for NDI Plus codes; for each study nosologist's original codes; and for "final study codes," derived by comparing the two sets of study nosologists' codes and resolving discrepancies by using the NCHS code. For all causes of death combined, the discrepancy rate was 4% for NDI Plus codes, 4% for the final study codes and 6%-7% for the study nosologists' original codes. The discrepancy rate for selecting the appropriate cancer site was 1% for NDI Plus codes and 3% for the final study codes. For noncancer conditions, the discrepancy rate was 5% for NDI Plus codes and 4% for the final study codes. NDI Plus underlying cause of death codes are comparable to codes developed using standard but more cumbersome procedures. The use of NDI Plus codes may enhance the validity of comparisons of an occupational cohort's mortality rates with national or state rates.