Personalized event detection methods and related devices and systems

Abstract

Medical devices and related patient management systems and event detection methods are provided. An exemplary method of detecting events pertaining to operation of a medical device, such as an infusion device, involves obtaining measurements indicative of a condition in a body of a patient, determining statistics for an analysis interval based on the measurements, and determining an event probability associated with the analysis interval based on historical event data associated with the patient. An event detection model associated with the patient is obtained and applied to the statistics and the event probability to identify occurrence of the event using the event detection model, and in response, an indication of the event associated with the analysis interval is provided, for example, by tagging or marking data in a database, displaying graphical indicia of the event, or the like.

What is claimed is: 
     
         1 . A method of detecting events pertaining to operation of a medical device associated with a patient, the method comprising:
 obtaining, by a computing device via a network, a plurality of measurements indicative of a physiological condition of the patient during an analysis interval;   obtaining, by the computing device, a patient-specific event detection model that identifies one or more measurement statistics that are correlative to occurrence of an event for the patient;   determining, by the computing device, values for the one or more measurement statistics characterizing the physiological condition of the patient within the analysis interval based on the plurality of measurements;   determining, by the computing device, an event probability associated with the analysis interval based on historical event data associated with the patient;   determining, by the computing device, a metric indicative of a likelihood of the event based on the values for the one or more measurement statistics and the event probability using the patient-specific event detection model;   autonomously detecting, by the computing device, an occurrence of the event during the analysis interval based on the metric; and   in response to detecting the occurrence of the event, providing, by the computing device, an indication of the occurrence of the event associated with the analysis interval.   
     
     
         2 . The method of  claim 1 , wherein the one or more measurement statistics include at least one of a mean measurement value over the analysis interval, a standard deviation of the plurality of measurements during the analysis interval, a mean rate of change of the plurality of measurements on a sample-to-sample basis during the analysis interval, a standard deviation associated with the mean rate of change of the plurality of measurements, an absolute amplitude of the plurality of measurements during the analysis interval, and an amplitude difference between first and last measurements of the plurality of measurements during the analysis interval. 
     
     
         3 . The method of  claim 2 , wherein the plurality of measurements comprises a plurality of glucose measurements indicative of a glucose level in a body of the patient. 
     
     
         4 . The method of  claim 1 , wherein the plurality of measurements comprises a plurality of glucose measurement values indicative of a glucose level in a body of the patient. 
     
     
         5 . The method of  claim 4 , wherein the one or more measurement statistics include at least one of a mean sensor glucose measurement value over the analysis interval, a standard deviation of the plurality of glucose measurement values during the analysis interval, a mean rate of change of the plurality of glucose measurement values on a sample-to-sample basis during the analysis interval, a standard deviation associated with the mean rate of change of the plurality of glucose measurement values, an absolute amplitude of the plurality of glucose measurement values during the analysis interval, and an amplitude difference between first and last glucose measurement values of the plurality of glucose measurement values during the analysis interval. 
     
     
         6 . The method of  claim 1 , further comprising:
 obtaining historical measurement data for the physiological condition of the patient;   obtaining historical event data for the patient; and   generating the patient-specific event detection model based on the historical measurement data and the historical event data.   
     
     
         7 . The method of  claim 6 , the patient-specific event detection model comprising an equation for calculating the likelihood of the event based on the values for the one or more measurement statistics and the event probability using respective correlation coefficients, wherein generating the patient-specific event detection model comprises:
 determining a plurality of measurement statistics based on the historical measurement data;   determining a predictive subset of the plurality of measurement statistics for the patient based on a respective correlation between a respective measurement statistic of the plurality of measurement statistics and the historical event data, the predictive subset comprising the one or more measurement statistics; and   determining the respective correlation coefficients associated with respective ones of the predictive subset, wherein the respective correlation coefficients weight the relative predictiveness or correlative strength of each respective measurement statistic of the one or more measurement statistics.   
     
     
         8 . The method of  claim 7 , wherein the physiological condition comprises a glucose level and the plurality of measurements comprises a plurality of glucose measurements. 
     
     
         9 . The method of  claim 7 , wherein determining the predictive subset comprises utilizing stepwise logistic regression to determine what measurement statistics among the plurality of measurement statistics are predictive of the event for the patient. 
     
     
         10 . The method of  claim 1 , wherein:
 the patient-specific event detection model comprises an equation for calculating the likelihood of the event based on model inputs using respective correlation coefficients;   the model inputs comprise the one or more measurement statistics and the event probability; and   the respective correlation coefficients weight the relative predictiveness or correlative strength of each respective input of the model inputs to occurrence of the event.   
     
     
         11 . The method of  claim 1 , wherein providing the indication comprises tagging the plurality of measurements prior to uploading the plurality of measurements to a server. 
     
     
         12 . The method of  claim 1 , the medical device comprising an infusion device, wherein providing the indication comprises automatically adjusting autonomous operation of the infusion device to temporarily modify delivery to account for the occurrence of the event. 
     
     
         13 . The method of  claim 1 , further comprising determining a timestamp for the occurrence of the event based on the plurality of measurements, wherein providing the indication comprises storing an event indicator having the timestamp associated therewith. 
     
     
         14 . The method of  claim 1 , wherein providing the indication comprises providing a pattern guidance display for an event pattern that is influenced by the occurrence of the event associated with the analysis interval. 
     
     
         15 . A computer-readable medium having instructions stored thereon that are executable by a processing system of the computing device to perform the method of  claim 1 . 
     
     
         16 . A method of detecting events during to operation of an infusion device associated with a patient, the method comprising:
 obtaining sensor measurements for a physiological condition of the patient;   obtaining historical event data for the patient;   obtaining an event detection model associated with the patient, the event detection model identifying a predictive subset of measurement statistics correlative to an event by the patient;   calculating the predictive subset of measurement statistics characterizing the physiological condition of the patient for an analysis interval of the sensor measurements;   determining an event probability associated with the analysis interval based on the historical event data;   determining a metric indicative of a likelihood of occurrence of the event based on the calculated predictive subset of measurement statistics for the analysis interval of the sensor measurements and the event probability associated with the analysis interval using respective correlation coefficient values from the event detection model;   autonomously detecting an occurrence of the event during the analysis interval based on the metric; and   in response to autonomously detecting the occurrence of the event, providing an indication of the occurrence of the event associated with the analysis interval.   
     
     
         17 . The method of  claim 16 , wherein providing the indication comprises updating the historical event data to indicate the occurrence of the event associated with the analysis interval. 
     
     
         18 . The method of  claim 16 , further comprising determining the event detection model associated with the patient based on historical sensor measurements for the patient corresponding to the historical event data. 
     
     
         19 . The method of  claim 18 , wherein determining the event detection model comprises:
 classifying the historical sensor measurements into event segments and non-event segments; and   determining a predictive measurement statistic of the predictive subset based on a correlation between the event segments and values of the predictive measurement statistic for portions of the historical sensor measurements corresponding to the event segments.   
     
     
         20 . A system comprising:
 a sensing arrangement to obtain measurement values for a physiological condition in a body of a patient;   a database to store historical event data and an event detection model associated with the patient; and   a computing device communicatively coupled to the database and a network to:
 obtain the measurement values; 
 determine values for one or more measurement statistics characterizing the physiological condition in the body of the patient that are correlative to occurrence of an event for the patient for an analysis interval based on the measurement values; 
 determine an event probability associated with the analysis interval based on the historical event data associated with the patient; 
 apply the event detection model to the one or more measurement statistics and the event probability to determine a metric indicative of a likelihood of occurrence of the event based on the values for the one or more measurement statistics and the event probability using respective correlation coefficient values from the event detection model and autonomously detect an occurrence of the event during the analysis interval based on the metric; and 
 provide an indication of the occurrence of the event associated with the analysis interval.