Data Entry

Each pt will be assigned a unique study ID. We also will enter the UMMC ID into the study record.  For the purposes of follow-up, we also collect data on name, name at birth, current address, day and evening telephone numbers, alternate contact name, and social security number. All data that identify an individual pt will be removed from analytic datasets.  All datasets with identifiers are password-protected, and are kept locked in the files room at the Prevention Institute (see Resources and Environment).

Quality Control Measures.

All CATI-derived data are entered directly into an NDS database, and then downloaded into SAS data sets.  Opscan forms are checked prior to scanning for obvious errors and stray marks.  All such checking is blinded as to the identity and study condition of the participant. All data from pencil and paper forms are entered using a double entry format. All of the data entry systems employ automatic checks for values that are out of range or represent errors of faulty logic. These procedures reduce transcription errors for hard copy data to close to zero (<0.5%). Therefore, for telephone interviews we will monitor a random 1% of all 24HR calls.  As we currently do for all data collected from transcription error-free methods such as the 24HR, we will conduct frequent exploratory analyses to detect outliers and to assess for adherence to the assumptions of the multivariable analyses, particularly regression modeling, that we plan to use in order to test hypotheses and to measure effect  (e.g., normality, linearity, and homoscedasticity). (129)

Management of Data Files.

In close collaboration with Drs. Hebert and Stanek, the project's Masters level biostatistician, T. Hurley, will develop the interactive database systems, write the appropriate software programs, and be responsible for quality control measures. Dr. Hebert will have primary responsibility for overseeing the assembling of datasets from pts in the study. Dr. Stanek, in collaboration with Dr. Hebert and Mr. Hurley, will have primary responsibility for structuring primary data sets, code books, naming conventions, and for designing primary documentation of data. Data management will be conducted in the Division of Preventive and Behavioral Medicine in a microcomputer environment.  Primary data sets will be shared via FTP with Dr. Stanek at the Umass-Amherst site. Data sets are cleaned, verified and archived, and then read into SAS data sets which also are archived. Access to programs and archival data sets will be denied by password to non-study staff.  All analytic and tracking database files are backed up daily.  On a weekly, monthly, quarterly and yearly basis complete backups are made of all database files.  One copy is saved on-site and one off-site. Separate archival databases are permanently maintained.