Heart failure detection with a sequential classifier

A system for detecting heart failure in a patient, comprising:

• a plurality of sensors configured to collect sensor measurements of physiological data from a patient;
• a processor configured to:
• calculate a first composite value using sensor measurements from at least two sensors of the plurality of sensors;
• and
• calculate a second composite value using sensor measurements from at least a third sensor of the plurality of sensors if the first composite value exceeds a first threshold or is outside of a first specified range, wherein the third sensor is different than the at least two sensors of the plurality of sensors used to calculate the first composite value;
• and
• a heart failure detection module configured to generate a worsening heart failure indicator that indicates a likelihood that the patient's heart failure status has worsened using at least the second composite value,
• wherein the processor is configured to calculate the first composite value by either subtracting a first long-term average of the sensor measurements from the at least two sensors from a first short-term average of the sensor measurements from the at least two sensors, or subtracting a first baseline value of the sensor measurements from the at least two sensors from the first long-term average of the sensor measurements from the at least two sensors, wherein the first short-term average is an average over a first plurality of days and the first long-term average is an average over a second plurality of days different from the first plurality of days, and
• wherein the processor is configured to calculate the second composite value by either subtracting a second long-term average of the sensor measurements from the at least the third sensor from a second short-term average of the sensor measurements from the at least the third sensor, or subtracting a second baseline value of the sensor measurements from the at least the third sensor from the second long-term average of the sensor measurements from the at least the third sensor, wherein the second short-term average is an average over a third plurality of days and the second long-term average is an average over a fourth plurality of days different from the third plurality of days.

A method of detecting heart failure in a patient via a medical system, comprising:

• collecting sensor measurements of physiological data from a patient using a plurality of sensors, wherein at least one of the plurality of sensors resides on a cardiac rhythm management device;
• calculating, with a processor within the medical system, a first composite value using sensor measurements from at least two sensors of the plurality of sensors, wherein the first composite value is an indication of a likelihood that the patient's heart failure status has changed;
• comparing the first composite value to a first threshold or a first specified range;
• if the first composite value exceeds the first threshold or is outside of the first specified range, then calculating a second composite value using sensor measurements from at least a third sensor of the plurality of sensors, wherein the third sensor is different than the at least two sensors of the plurality of sensors used to calculate the first composite value, wherein the second composite value is an indication of a likelihood that the patient's heart failure status has changed;
• comparing the second composite value to a second threshold or a second specified range;
• and
• if the second composite value exceeds the second threshold or is outside of the second specified range, generating a worsening heart failure indicator via the processor within the medical system, the worsening heart failure indicator indicating a likelihood that the patient's heart failure status has worsened;
• wherein calculating the first composite value includes either subtracting a first long-term average of the sensor measurements from the at least two sensors from a first short-term average of the sensor measurements from the at least two sensors, or subtracting a first baseline value of the sensor measurements from the at least two sensors from the first long-term average of the sensor measurements from the at least two sensors, wherein the first short-term average is an average over a first plurality of days and the first long-term average is an average over a second plurality of days different from the first plurality of days, and
• wherein calculating the second composite value includes either subtracting a second long term average of the sensor measurements from the at least the third sensor from a second short-term average of the sensor measurements from the at least the third sensor, or subtracting a second baseline value of the sensor measurements from the at least the third sensor from the second long term average of the sensor measurements from the at least the third sensor, wherein the second short-term average is an average over a third plurality of days and the second long-term average is an average over a fourth plurality of days different from the third plurality of days.