Hawkins B S, Singer S W
Control Clin Trials. 1986 Jun;7(2):89-117. doi: 10.1016/0197-2456(86)90027-9.
We were given the opportunity to design and implement a general data processing system to accommodate several different epidemiologic studies to be conducted by a new research group. A survey of 15 operating data centers was conducted in preparation for undertaking the design and development of our system. The results of the survey indicated that data processing activities can be classified, both conceptually and operationally, into three modules: data recording and data entry, data management, and data analysis, and that the data management functions were those amenable to generalization. Based on our survey and the varying needs of our studies, we selected a "mixed" hardware environment, using both a computer center mainframe and microcomputers. We created the systems using commercially available software, including a mainframe database manager and mainframe statistics packages, microcomputer data entry software, and a communications package to link the two environments. Our strategy was to buy software, when possible, rather than to build custom programs, and to let software tools govern hardware needs. Hardware independence, price, and functional capability directed our software choices, while hardware selection was constrained most importantly by available software, then by budget, by available computing resources, and finally by the marketplace. The system has been used successfully in three studies differing in design, size, data collection locale, and rate of data accrual.
我们获得了一个机会，来设计并实施一个通用数据处理系统，以适应一个新研究团队即将开展的多项不同的流行病学研究。为准备我们系统的设计与开发，我们对15个运营数据中心进行了调查。调查结果表明，数据处理活动在概念和操作上都可分为三个模块：数据记录与数据录入、数据管理以及数据分析，并且数据管理功能适合进行通用化处理。基于我们的调查以及各项研究的不同需求，我们选择了一种“混合”硬件环境，同时使用计算机中心的大型机和微型计算机。我们使用商业上可用的软件创建了这些系统，包括大型机数据库管理器和大型机统计软件包、微型计算机数据录入软件以及一个用于连接这两种环境的通信软件包。我们的策略是尽可能购买软件，而非构建定制程序，并让软件工具决定硬件需求。硬件独立性、价格和功能能力指导了我们对软件的选择，而硬件选择最重要的限制因素是可用软件，其次是预算、可用计算资源，最后是市场情况。该系统已成功应用于三项在设计、规模、数据收集地点以及数据积累速度方面存在差异的研究中。
