|
|
|
Andreas Seyfang and Silvia Miksch |
|
Vienna University of Technology |
|
Austria |
|
|
|
|
|
Motivation |
|
Why Guideline Execution |
|
Why Temporal Data Abstraction |
|
Definitions |
|
Abstractions with example |
|
Strength & limitations |
|
|
|
|
|
Deliver the right recommendation at the right
time |
|
Reduce information overload |
|
Improve quality of health care |
|
Prerequisites |
|
Information about patient state |
|
Formal representation of guideline |
|
|
|
|
|
Integration into clinical data flow necessary |
|
Additional data entry
= additional work
= barrier to usage of
guideline execution system |
|
|
|
|
|
|
|
|
Integration into clinical data flow necessary |
|
Gap between raw data and medical concepts |
|
Quantitative raw data: |
|
11:23:05 SpO2=96% |
|
11:23:06 SpO2=95% |
|
11:23:07 SpO2=96% |
|
Qualitative medical concept:
sufficient
oxygen supply in artificial ventilation |
|
|
|
|
|
Temporal dimension crucial part of medical
concepts (often implicit) |
|
Recent readings of SpO2 |
|
Combinations of different time windows necessary |
|
Short term trend can invalidate
long term observation |
|
|
|
|
|
|
Parameter = Variable plus history of
measurements |
|
Episode = period of time during which a
parameter has a certain value |
|
Parameter proposition
= parameter
+ value constraint
+ context
+ temporal constraints |
|
|
|
|
|
Solution |
|
Parameter proposition |
|
Parameter name: SpO2 |
|
Value constraint: less than 80 |
|
Context: artificial ventilation |
|
Minimum duration: 4 seconds |
|
|
|
|
Qualitative values |
|
Sliding time windows |
|
Statistical measures |
|
Time/date oriented abstraction |
|
Repetitions |
|
|
|
|
|
Logical |
|
and, or |
|
Arithmetic |
|
sum, difference |
|
|
|
Different parameter propositions |
|
Aggregates of different time ranges |
|
|
|
|
|
Solution |
|
Hypoxy as before |
|
Hyperoxy similar but time constraint:
Latest start 20 seconds after end of hypoxy |
|
|
|
|
|
|
Detection of episodes |
|
|
|
Aggregates for sliding time windows |
|
|
|
Monitoring of repetitions |
|
|
|
Free combination of abstractions |
|
|
|
|
Knowledge acquisition |
|
Access to all required inputs |
|
Integration with precise formalized guideline |
|
|
|
Some abstractions not implemented |
|
|
|
|
|
|
Guideline execution needs temporal data
abstraction |
|
|
|
Temporal data abstraction needs guideline
execution |
|
|
|
Notes