An annotated data collection system to support intelligent analysis of Intensive Care Unit data

Abstract

Without a proper system for collecting data, the extent to which intelligent analysis can be performed on that data is limited. This is especially true in the study of time-ordered data, in which the gold standard for analysis techniques often comes from knowledge of discrete event occurrences amongst continuous data streams. An illustrative example is the study of patient monitor data from the Intensive Care Unit (ICU), which holds the promise of improving patient care in a setting where data overload and false alarms currently make that goal difficult at best. Here, monitor data is practically useless without corresponding knowledge of what clinical events were taking place to produce the observed data. A method of collecting both monitor data and clinical event annotations, and subsequently being able to correlate the two, has been developed and is being used at Children's Hospital in Boston. Preliminary results indicate that this type of data collection system is a viable tool for facilitating the intelligent analysis of temporal data, such as that from the ICU.