Inaccurate data entry can be costly. Without accurate data, orders can't be filled correctly and market forecasting can run amok. At Isis Pharmaceuticals, a $26 million drug development company in Carlsbad, Calif., database records are routinely reviewed by the Food and Drug Administration, which could bring operations to a halt if a few figures are entered incorrectly. So Isis adheres religiously to a total quality system for data entry that keeps the database as accurate as possible. Here are the five steps in the process.

  1. Manual review. Clinical-data coordinator checks entry manually for obvious mistakes.
  2. Double entry. Once an individual's information has been punched into the database, a second data entry operator reenters the same information in a new file. The computer then automatically compares the two files and prints out a list of inconsistencies, which must be reconciled by the clinical-data coordinator.
  3. Visual verification. A data manager compares the data that were entered against the original handwritten forms.
  4. Electronic verification. A data manager runs a series of relational edit-check programs that examine data in selected fields within the database to make sure they are logical and consistent.
  5. Final audit. Once an entire clinical trial is completed, Stuart Nixon, director of clinical-data management and biostatistics, and his team conduct one final audit by checking a representative cross-section of data against the original handwritten forms.