Browse Prior Art Database

METHOD FOR SMOOTHING VARIABLE DIAGNOSTICS TO ASSESS PATIENT STATUS

IP.com Disclosure Number: IPCOM000130348D
Publication Date: 2005-Oct-21
Document File: 3 page(s) / 48K

Publishing Venue

The IP.com Prior Art Database

Abstract

A method of analyzing patient diagnostics including measuring patient data, and performing a weighted linear regression. The method is useful to reduce noise content in the diagnostics. For diagnostics recorded on a daily basis, an n-day weighted linear regression over a duration long enough to average out uninteresting variability (n=28 days for example) is used to create a less variable trend for analysis of patient health. The days are weighted according to the formula: weight= e(D/C), where the current day is D=0, yesterday is D= -1, 2 days ago is D= -2, and so on and C is a constant (C=10 for example). Weighted values are used in a linear regression to create a linear trend line using the equation y=mD+b. This method is repeated each day to yield an expected value for the current day (b) based on the actual value recorded by the device (D0), and also based on how this parameter has been behaving over the past n days (D1 through Dn). Thus the values calculated from this linear fit are less subject to environmental and physiologic "noise" than the recorded daily values.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 52% of the total text.

Method for smoothing variable diagnostics

to assess patient status

Recorded device-based and external patient diagnostics tend to vary over time.  Variations depend on multiple factors including timing, drug use, and physiologic and environmental changes.  Uninteresting variations in these measurements are termed “noise”.  Caregivers seek diagnostic information which has low noise content.  The present subject matter provides methods for manipulating diagnostic information to control noise content.

For diagnostics recorded on a daily basis, a n-day weighted linear regression over duration long enough to average out uninteresting variability (n=28 days for example) is used to create a less variable trend for analysis of patient health.  The days are weighted according to the formula: weight= e(D/C), where the current day is D=0, yesterday is D= -1, 2 days ago is D= -2, and so on and C is a constant (C=10 for example).  Weighted values are used in a linear regression to create a linear trend line using the equation y=mD+b.  This method is repeated each day to yield an expected value for the current day (b) based on the actual value recorded by the device (D0), and also based on how this parameter has been behaving over the past n days (D1 through Dn).  Thus the values calculated from this linear fit are less subject to environmental and physiologic "noise" than the recorded daily values.

For diagnostics that are measured semi-daily or weekly, this same method can be applied using different, optimized time constants in the weight and linear fits based on a larger time window.  For example if a particular diagnostic is recorded once a week, the linear fit may be based on n=56 days rather than n=28, and the time constant in the weight may be C=20 instead of C=10 (weight= e(D/20)). 

Fig. 1...