The use of teratogen information services for research: Assessment of reliability of data entry

Teratogen information services (TIS) have been able to prospectively follow large cohorts of women with specific exposures during pregnancy. TIS researchers have increasingly relied on computerized databases to retrieve information, although little attention has been focused on the quality of the actual data. We examined the reliability of data entry at Motherisk, one such TIS. Data abstracted from 37 semi-standardized clinic forms were entered into the database by three counselors experienced in data entry. Inter- and intracoder reliability were assessed by intraclass correlation coefficient (ICC) for continuous variables and kappa (K) or weighted kappa (K_w) for categoricalvariables. An ICC, K, or K_w value of ≥0.60 with a P < 0.05 was considered adequate. Demographics, obstetric history, smoking, and caffeine intake were all reliably entered. The number of primary exposures and alcohol were not. The former stemmed from the unreliable data entry of one coder, whereas the latter resulted from misuse of the alcohol nomogram by all coders. As a result, data entry has now been assigned to one person (who will be assessed for intrarater reliability), and our quality assurance program stresses staff retraining in all aspects of data collection.