Pivot/Remote: a distributed database for remote data entry in multi-center clinical trials.
1. INTRODUCTION. Data collection is a critical component of multi-center clinical trials. Clinical trials conducted in intensive care units (ICU) are even more difficult because the acute nature of illnesses in ICU settings requires that masses of data be collected in a short time. More than a thousand data points are routinely collected for each study patient. The majority of clinical trials are still "paper-based," even if a remote data entry (RDE) system is utilized. The typical RDE system consists of a computer housed in the CC office and connected by modem to a centralized data coordinating center (DCC). Study data must first be recorded on a paper case report form (CRF), transcribed into the RDE system, and transmitted to the DCC. This approach requires additional monitoring since both the paper CRF and study database must be verified. The paper-based RDE system cannot take full advantage of automatic data checking routines. Much of the effort (and expense) of a clinical trial is ensuring that study data matches the original patient data. 2. METHODS. We have developed an RDE system, Pivot/Remote, that eliminates the need for paper-based CRFs. It creates an innovative, distributed database. The database resides partially at the study clinical centers (CC) and at the DCC. Pivot/Remote is descended from technology introduced with Pivot [1]. Study data is collected at the bedside with laptop computers. A graphical user interface (GUI) allows the display of electronic CRFs that closely mimic the normal paper-based forms. Data entry time is the same as for paper CRFs. Pull-down menus, displaying the possible responses, simplify the process of entering data. Edit checks are performed on most data items. For example, entered dates must conform to some temporal logic imposed by the study. Data must conform to some acceptable range of values. Calculations, such as computing the subject's age or the APACHE II score, are automatically made as the data is entered. Data that is collected serially (BP, HR, etc.) can be displayed graphically in a trend form along with other related variables. An audit trail is created that automatically tracks all changes to the original data, making it possible to reconstruct the CRF to any point in time. On-line help provides information on the study protocol as well as assistance with the use of the system. Electronic security makes it possible to lock certain parts of the CRF once it has been monitored. Completed CRFs are transmitted to the DCC via electronic mail where it is reviewed and merged into the study database. Questions about subject data are transmitted back to the CC via electronic mail. This approach to maintaining the study database is unique in that the study data files are distributed among the CC and DCC. Until a subject's CRF is monitored (verified against the original patient data residing in the hospital record), it logically resides at the CC where it was collected. Copies are transmitted to the DCC and are only read there. Any pre-monitoring changes must be made to the data at the CC. Once the subject's CRF is monitored, it logically moves to the DCC, and any subsequent changes are made at the DCC with copies of the CRF flowing back to the CC. 3. DISCUSSION. Pivot/Remote eliminates the need for paper forms by utilizing portable computers that can be used at the patient bedside. A GUI makes it possible to quickly enter data. Because the user gets instant feedback on possible error conditions, time is saved because the original data is close at hand. The ability to display trended data or variables in the context of other data allows detection of erroneous conditions beyond simple range checks. The logical construction of the database minimizes the problem of managing dual databases (at the CC and DCC) and keeps CC personnel in the loop until all changes are made.