Signal Data
Data Description
Various types of signal data has been included in the SICDb dataset. Signal data is saved in data_float_h table, value reference is found in d_references.
Important note: Data stored in data_float_h is aggregated hourly data.
Predefined Fields (RooDataServer)
| Name | Description | Comment | Type |
|---|---|---|---|
| SignalCountHours | Amount of hours where selected signal data is available | O D | |
| SignalAvg | Average signal value of selected signal | O D | |
| SignalMin | Minimum hourly average signal value of selected signal | O D | |
| SignalMax | Maximum hourly average signal value of selected signal | O D | |
| SignalHoursUnderX | Amount of hours, where the average signal value is under X | O D | |
| SignalHoursOverX | Amount of hours, where the average signal value is >= X | O D | |
| SignalHoursBetweenXY | Amount of hours, where the average signal value is >=Min and <Max | O D | |
| SignalStd | Standard deviation of average hourly values of selected signal | O D
|
Aggregation Description
Most signal data is averaged. The field data_float_h.cnt describes the amount of values aggregated. It is to be expected that the first signal values have less than 60 data points, as it is quite unlikely (1:60) that signal starts on minute 0.
Raw values are saved in field data_float_h.rawdata, being a list of bytes. (IEEE 754 Single(4 bytes), LittleEndian)
Metavision does, in database, not differentiate between qualitative (i.e. blood pressure) and quantitative (i.e. drainage volume) data, this was done programatically saving the sum instead of average.
Known Limitations
0 Values
All values were aggregated to hourly average values. As in many cases Metavision saves 0 values during signal loss times instead of NULL aggregating these would significantly influence statistics. Think about a patient with constant MAP of 60 is disconnected for 15min, aggregating would result in a average MAP of 45. For this reason 0 values were set to NULL and ignored. This behaviour is expected to represent in nearly all cases the most likely value.
In very special cases, like cardioplegia, where 0 would be a correct pulse rate, expect values to be missing. Contact developer in case you need this data.