MORE INNOVATIVE STRATEGIES NEEDED TO ACHIEVE THE GOAL OF TUBERCULOSIS ELIMINATION

My colleagues and I thank Rosenberg for hiscomments and interpretation regarding ourwork on the deterioration of American Indianmortality statistics in Maine. Rosenberg citesimprovements made by the National Centerfor Health Statistics (NCHS) in processing pro-cedures for race-specific data. Undoubtedlythese new procedures have improved theoverall quality of race coding. However, theseprocedures involve the systematic review andediting of electronic records, which have al-ready undergone data entry. These proce-dures would not address the problem associ-ated with data entry that we documented.The gold standard for detecting and resolv-ing data entry errors is double data entry.Historically, in Maine, all vital records datawere double entered, the 2 files were com-pared, and any discrepancies were resolved.In the early 1990s, Maine death certificatedata processing was restructured because ofreductions in the Maine State General Fund;double data entry was discontinued. It istherefore not surprising that we found nodata entry errors before 1988.NCHS randomly selects a 10% sample ofMaine’s records; these are reentered and com-pared with the original electronic files. In stateswith small non-White populations, this proce-dure may not identify systematic race-specificcoding errors. One approach for validating theAmerican Indian race is to select all deathsamong residents of towns in or adjacent toAmerican Indian reservations. However, thisapproach would capture information only aboutAmerican Indians residing on reservations.While using rosters of American Indianpopulations has been successful (IndianHealth Service rosters,