Many pediatric critical-care units maintain patient databases that are used for a variety of activities that include research, billing, planning, and documenting resource utilization. Increasing interest in outcomes analysis and quality assurance and improvement activities are other driving forces for data collection in these settings. The Division of Critical Care Medicine and the General Clinical Research Center at the Cornell University Medical Center, and the Pediatric Critical Care Study Group (PCCSG) collaborated in creating a database to facilitate multi-institutional research and to help define the scope of pediatric critical care.

The purpose of the database was to promote research by providing the capability to identify cases of interest and to monitor global trends. The goal of this collaboration was to standardize data collection across many PICUs to allow the integration of local data into a centralized master database for query and analysis. A distributed system of flexible and multiply-compatible programs was to be devised to run on personal computers. With these programs pediatric intensivists would collect demographic and clinical data to upload to a central registry. The collection instrument was to be flexible enough for individual institutions to add data elements and for research modules to be added as multi-institutional research protocols developed. The data were to be stored in a format that could be queried locally by the database management software with which the users were most familiar.

Several other considerations were to be addressed by the implementation. It was also recognized that the value of the data must exceed the tedium of collecting it. To this end the data elements were stripped to a minimum core that would interest virtually every participating institution. Data coding libraries were tediously compiled for retrievability, consistency, and in some cases, confidentiality. These libraries were based on experience with our own database and on the interests of pediatric intensivists. Confidentiality an imperative, for both patients and institutions was also considered in the design. Several features, including password protection and data stripping or encryption, have been implemented to safeguard data confidentiality. Back up and archiving features were added to secure data locally. Different data integrity problems problems were addressed with cross-checking algorithms and carefully planned human engineering to maximize accuracy of data ! input.

In 1991, the first version of the PCCSG Data Acquisition Program, the PCCSGNet©, was distributed. It utilized the dBase file format because of its wide-spread use and Fox software because of its portability to both the PC and Macintosh platforms. The initial plan was to deliver the software to a few beta-test sites, but demand was so great that we provided the beta version to 70 institutions. With much valuable user the data acquisition program was further refined, and release 2.0 began shipment in the fall of 1994. Among the major changes were: migration to the more powerful FoxPro engine, an improved user interface, enhanced coding libraries and utilities, improved security, augmented flexibility, and significantly enhanced reporting functions. Significantly improved documentation, both on-line and in a printed manual were also made available.

Although there are continuing challenges, we have made tremendous progress in developing consistent data collection among many PICUs across the country. Software has been distributed to over 50 institutions under a shareware agreement. The nominal fee requested of registered users goes toward defraying some of the development and maintenance costs. Registration entitles users to unlimited user support, free upgrades up until the next major revision, a copy of the printed documentation, and of course, authorization to use the program.

A discharge letter generator will soon be added to the set of reporting functions in the next interim upgrade. Future plans include rebuilding the foundations of the program to safeguard data integrity better and provide better performance. A facelift of the interface and an extensive revision of the coding library are also on the agenda, as is providing the users with improved, more flexible querying tools. With data uploaded from participating institutions, the Project will begin receiving proposals for queries toward the end of this year.

For more information, contact Steven Pon, MD (spon@mail.med.cornell.edu), at 212-746-3275 or pccsgnet@crcvax.med.cornell.edu