REGRESSION METHODOLOGY FOR THE EVALUATION OF A MEDICAL INFORMATION SYSTEM.

摘要

The information services provided by the data center of a clinical trial, observational study or registry include data collection, entry, editing, archiving, and analysis. One of the most important specific responsibilities of the data center is development of a data management system providing delivery of reliable data in a timely and efficient manner. This research offers a criterion for quantitative performance evaluation called the data processing sojourn. The sojourn is defined as the duration between an event of data accrual and acquisition of that data into the database or reporting system.;The activity of data management in any data center can be formulated in terms of quantification of the sojourn and the control of its mean and variance. This new criterion is simple, intuitive, and its implementation requires stringent project control. Regression methodology helps the practitioner to understand the performance of data management system and the results of hypothesis tests should be easily understood by the budget officers, the data management staff, and the principal investigators. This research indicates that quantification of the activity of data management can be a useful tool for guiding the design, management, evaluation, budgeting, comparison, even selection, of data management strategies.;A regression methodology is developed using the key assumptions that: the data submitted for processing are collected at the time of accrual; these data are processed in a fixed number of steps; one and only one step of processing occurs at any time; no steps are repeated; no processing follows acquisition. Regression covariates for this process are grouped into categories of: system design and operations policy; information quality; system resources; and (patient's) diagnostic variables. Operations data from the Patient Registry for Primary Pulmonary Hypertension are used as the basis for a case study.