| Autor: |
J Flynn A; School of Information, University of Michigan, Ann Arbor, MI.; Medical School, University of Michigan, Ann Arbor, MI., Milstein JA; School of Information, University of Michigan, Ann Arbor, MI.; School of Medicine, University of California San Francisco, San Francisco, CA., Boisvert P; Medical School, University of Michigan, Ann Arbor, MI., Gittlen N; Medical School, University of Michigan, Ann Arbor, MI., Lagoze C; School of Information, University of Michigan, Ann Arbor, MI., Meng G; Medical School, University of Michigan, Ann Arbor, MI. |
| Abstrakt: |
Many obstacles must be overcome to generate new biomedical knowledge from real-world data and then directly apply the newly generated knowledge for decision support. Attempts to bridge the processes of data analysis and technical implementation of analytic results reveal a number of gaps. As one example, the knowledge format used to communicate results from data analysis often differs from the knowledge format required by systems to compute advice. We asked whether a shared format could be used by both processes. To address this question, we developed a data-to-advice pipeline called ScriptNumerate. ScriptNumerate analyzes historical e-prescription data and communicates its results in a compound digital object format. ScriptNumerate then uses these same compound digital objects to compute its advice about whether new e-prescriptions have common, rare, or unprecedented instructions. ScriptNumerate demonstrates that data-to-advice pipelines are feasible. In the future, data-to-advice pipelines similar to ScriptNumerate may help support Learning Health Systems. |
