Methods and systems for adverse event prediction using pump operating data

Abstract

The present disclosure provides for each of a method of characterizing divergence of a monitored flow rate of blood through an implantable blood pump, and a method of predicting an upcoming adverse cardiac event using operating data of the blood pump, such as the characterized divergence data. Predicting an upcoming adverse cardiac event may be further based on similar operating data from a plurality of other implantable blood pumps.

What is claimed is: 
     
         1 . A method of predicting an upcoming adverse event for a user of a first implantable blood pump based on information derived from first operating data of the first implantable blood pump, and further based on information derived from segments of second operating data of the first implantable blood pump and/or one or more other implantable blood pumps for which it is known whether each segment corresponds to occurrence or non-occurrence of an adverse event, wherein for each implantable blood pump, said operating data is gathered over a duration of time, and wherein the information is derived from a waveform constructed based on the operating data, the method comprising:
 (i) acquiring the first operating data from the first implantable blood pump,   (ii) deriving information from a waveform constructed based on the operating data,   (iii) accessing information derived from a waveform constructed based on the segments of second operating data (from the first or other implantable pumps) for which at least a first portion of the information corresponds to non-occurrence of an adverse event and at least a second portion of the information corresponds to occurrence of an adverse event,   (iv) for at least two waveform features of the derived information, comparing said information derived from the first operating data to said information derived from the second operating data, and   (v) based on the comparison, determining whether the first operating data correlates with the first or second portions of the second operating data, wherein a correlation to the second portion of the second operating data is a prediction of an upcoming adverse event within the predetermined time span.   
     
     
         2 . The method of predicting an upcoming adverse event as recited in  claim 1 , wherein said adverse event comprises one or more of a predicted bleeding, infection, cardiovascular health condition, cerebrovascular adverse event, hematocrit changes, and VAD peripheral malfunction. 
     
     
         3 . The method of predicting an upcoming adverse event as recited in  claim 1 , wherein said operating data for each implantable blood pump is a flow rate of blood in said implantable blood pump. 
     
     
         4 . The method of predicting an upcoming adverse event as recited in  claim 1 , wherein the predetermined time span is between about six days and about sixty days. 
     
     
         5 . The method of predicting an upcoming adverse event as recited in  claim 1 , wherein for each blood pump for which segments of second operating data are accessed, the segments of second operating data of said blood pump correspond to at least about one month of operating data. 
     
     
         6 . The method of predicting an upcoming adverse event as recited in  claim 1 , wherein the segments of second operating data are accessed at least in part from the first implantable blood pump, and wherein the segments of second operating data accessed from the first implantable blood pump are logged at least about one year prior to being accessed. 
     
     
         7 . The method of predicting an upcoming adverse event as recited in  claim 1 , wherein the second operating data comprises more than two portions, and wherein at least one portion of the second operating data corresponds to a particular type of adverse event. 
     
     
         8 . The method of predicting an upcoming adverse event as recited in  claim 1 , wherein at least one of said waveform features is selected from the group consisting of:
 an integral between two consecutive zero-crossing nodes, an integral between two non-consecutive zero-crossing nodes; a maximum calculated divergence of flow;   a minimum calculated divergence of flow;   a number of positive peaks over a time segment between two consecutive zero-crossing nodes;   a number of positive peaks over a time segment between two consecutive zero-crossing nodes;   a frequency of positive peaks between two consecutive zero-crossing nodes;   a frequency of negative peaks between two consecutive zero-crossing nodes;   a slope between a zero-crossing nodes and a positive peak;   a slope between a zero-crossing nodes and a negative peak; and   a slope between local extrema.   
     
     
         9 . The method of predicting an upcoming adverse event as recited in  claim 8 , wherein at least two of said waveform features are selected from the group of waveform features recited in  claim 8 . 
     
     
         10 . The method of predicting an upcoming adverse event as recited in  claim 1 , wherein said comparison is performed using a data mining analysis, wherein said data mining analysis is one of a linear discriminant analysis, a cluster analysis, and a Bayesian analysis. 
     
     
         11 . The method of predicting an upcoming adverse event as recited in  claim 1 , wherein said comparison is performed using a neural network. 
     
     
         12 . The method of predicting an upcoming adverse event as recited in  claim 1 , wherein the method further comprises determining a percentage likelihood of said prediction of an upcoming adverse event within the predetermined time span. 
     
     
         13 . The method of predicting an upcoming adverse event as recited in  claim 1 , wherein, in the event of a prediction of an upcoming adverse event within the predetermined time span, the method further comprises outputting a notification of said prediction and a percentage likelihood of said prediction occurring. 
     
     
         14 . The method of predicting an upcoming adverse event as recited in  claim 13 , further comprising outputting an analytical basis for the prediction, wherein the analytical basis for the prediction comprises an indication of whether second operating data that correlates with the first operating data was accessed from said first implantable blood pump or from said one or more other implantable blood pumps. 
     
     
         15 . The method of predicting an upcoming adverse event as recited in  claim 1 , wherein the prediction of an upcoming adverse event within the predetermined time span has a specificity of about 99%. 
     
     
         16 . The method of predicting an upcoming adverse event as recited in  claim 1 , wherein the prediction of an upcoming adverse event within the predetermined time span has an area-under-curve of about 99%. 
     
     
         17 . The method of predicting an upcoming adverse event as recited in  claim 1 , further comprising accessing data relating to a physiological factor of a user of said first blood pump, and wherein the determination of whether the information derived from the first operating data of said first blood pump correlates with the information derived from the first or second portions of information derived from the second operating data is based further on the physiological factor data. 
     
     
         18 . The method of predicting an upcoming adverse event as recited in  claim 17 , wherein the physiological factor is one of an activity level, a hematocrit level, a medication of the user, a pathology of a condition of the user, and a previously recorded adverse event of the user. 
     
     
         19 . The method of detecting thrombosis in an implantable blood pump, the method comprising (i) repeatedly estimating a flow rate of blood through the pump over time, (ii) repeatedly calculating a moving flow rate average of the estimated flow rate, (iii) repeatedly calculating a divergence between a then-current estimated flow rate and the then-current moving flow rate average, (iv) repeatedly calculating a moving sum of the calculated divergence, and (v) determining the presence of thrombosis in the blood pump based on the calculated moving sum. 
     
     
         20 . The method of detecting thrombosis in an implantable blood pump as recited in  claim 19 , further comprising (vi) receiving a first data set of moving sum data during which thrombosis did not occur, (vii) receiving a second data set of moving sum data during which a thrombosis did occur, and (viii) comparing the calculated moving sum to each of the first and second data sets, wherein determining the presence of thrombosis in the blood pump based on the calculated moving sum is based on said comparison.