An architecture for linking medical decision-support applications to clinical databases and its evaluation
An architecture for linking medical decision-support applications to clinical databases and its evaluation
复制标题
DOI：
10.1016/j.jbi.2008.10.007
复制
发表时间：
2009-04-01
影响因子：
4.5
通讯作者：
Shahar, Yuval
中科院分区：
文献类型：
--
作者：
German, Efrat;Leibowitz, Akiva;Shahar, Yuval
我们描述并评估了一个框架，即医学数据库适配器（MEIDA），用于使用标准医学模式和词汇将基于知识的医疗决策支持系统（MDSS）与多个临床数据库连接起来。我们的解决方案涉及一组用于在 MDSS 知识库 (KB) 中嵌入标准术语和单元的工具；一组方法和工具，用于使用三种启发法（词汇表的选择、关键术语的选择和测量单位的选择）将本地数据库（DB）模式以及与 MDSS 知识库相关的术语和单位映射为标准化模式、术语和单位；以及一组工具，它们在运行时自动将源自知识库的标准术语查询映射到使用本地数据库的模式、术语和单位制定的查询。通过将三个 KB 映射到三个 DB，成功评估了该方法。即使在使用其他启发式方法之后，使用单元域匹配启发式方法也能将术语映射候选者的数量平均减少 71%。运行时访问 10,000 条记录需要一秒钟。我们的结论是，使用三阶段方法和几种术语映射启发法将 MDSS 映射到不同的本地临床数据库既可行又有效。 (C) 2009 年，爱思唯尔公司出版。
We describe and evaluate a framework, the Medical Database Adaptor (MEIDA), for linking knowledge-based medical decision-support systems (MDSSs) to multiple clinical databases, using standard medical schemata and vocabularies. Our solution involves a set of tools for embedding standard terms and units within knowledge bases (KBs) of MDSSs; a set of methods and tools for mapping the local database (DB) schema and the terms and units relevant to the KB of the MDSS into standardized schema, terms and units, using three heuristics (choice of a vocabulary, choice of a key term, and choice of a measurement unit); and a set of tools which, at runtime, automatically map standard term queries originating from the KB, to queries formulated using the local DB's schema, terms and units. The methodology was successfully evaluated by mapping three KBs to three DBs. Using a unit-domain matching heuristic reduced the number of term-mapping candidates by a mean of 71% even after other heuristics were used. Runtime access of 10,000 records required one second. We conclude that mapping MDSSs to different local clinical DBs, using the three-phase methodology and several term-mapping heuristics, is both feasible and efficient. (C) 2009 Published by Elsevier Inc.