PURPOSE: A Swedish National Cataract Register was instituted in 1992, monitoring nearly all cataract operations in Sweden, and since its inception comprising about 95% of all operations. Data from a total of approximately 290 000 operations have been collected during 1992-1998. Data quality is an important factor, and we have therefore assessed the various types and frequencies of data entry errors in the material. METHODS: The medical records for all operations in five selected participating clinics were retrieved for a set month. Each data transfer step from the record to the final data base was monitored for a total of 574 operations. A total of 10 variables were recorded for each operation. RESULTS: Significant sources of error were absent in most variables. However, possibly important errors appeared in three: "date entering waiting list", "preoperative best corrected visual acuity in the operated eye", or "visual acuity in the fellow eye". There were also noteworthy variations between the five clinics, different for different parameters. Errors were predominantly prone to appear at the very first step of registration. In most cases this was due to deviations from the data collection instructions. CONCLUSIONS: The reliability is good for most values entered into the register, but it is important to ensure that data definitions are exact and adhered to. Repeated information to the involved persons on how to fill in the forms appears to be a requisite for maintaining good input quality.
PURPOSE: A Swedish National Cataract Register was instituted in 1992, monitoring nearly all cataract operations in Sweden, and since its inception comprising about 95% of all operations. Data from a total of approximately 290 000 operations have been collected during 1992-1998. Data quality is an important factor, and we have therefore assessed the various types and frequencies of data entry errors in the material. METHODS: The medical records for all operations in five selected participating clinics were retrieved for a set month. Each data transfer step from the record to the final data base was monitored for a total of 574 operations. A total of 10 variables were recorded for each operation. RESULTS: Significant sources of error were absent in most variables. However, possibly important errors appeared in three: "date entering waiting list", "preoperative best corrected visual acuity in the operated eye", or "visual acuity in the fellow eye". There were also noteworthy variations between the five clinics, different for different parameters. Errors were predominantly prone to appear at the very first step of registration. In most cases this was due to deviations from the data collection instructions. CONCLUSIONS: The reliability is good for most values entered into the register, but it is important to ensure that data definitions are exact and adhered to. Repeated information to the involved persons on how to fill in the forms appears to be a requisite for maintaining good input quality.