Method and apparatus providing an ongoing and real time indicator for survival and major medical events

A device for determining a personalized circadian rhythm identification and prediction of a patient comprising:

• one or more wearable device(s) capable of obtaining or having obtained data from during one or more daytime and nighttime cycles;
• a processor and a machine learning algorithm that:
• uses an XGBoost algorithm modified for categorical analysis by splitting data into trees using a pre-computed number of categories based on first axis, second axis, and third axis measurements, which trees also include independent dummy variables to eliminate when no data is available, wherein the modified XGBoost algorithm generates a predicted trajectory of resting and active phase of when there will be little to no day, or night activity levels of the patient;
• determines a daytime threshold and a nighttime threshold from the data obtained during the one or more daytime and nighttime cycles by:
• identifying one or more daytime and nighttime boundaries by:
• selecting from at least one of physical activity, heart rate, or sleep status data;
• temporally segmenting the physical activity, heart rate, or sleep status data to establish a circadian rhythm;
• using at least one of a 2-mean, 5-mean, or K-nearest neighbor algorithm to determine the one or more daytime and nighttime boundaries;
• and
• using the processor to identify the personalized circadian rhythm.

A non-transitory computer readable medium for providing an ongoing and real time indicator for determining a personalized circadian rhythm identification and prediction from wearable data obtained from one or more wearable device(s) comprising:

• obtaining or having obtained data from the one or more wearable device(s) during one or more daytime and nighttime cycles;
• using a processor and a machine learning algorithm comprising:
• using an XGBoost algorithm modified for categorical analysis by splitting data into trees using a pre-computed number of categories based on first axis, second axis, and third axis measurements, which trees also include independent dummy variables to eliminate when no data is available, wherein the modified XGBoost algorithm generates a predicted trajectory of resting and active phase of when there will be little to no day, or night activity levels of a patient;
• determining a daytime threshold and a nighttime threshold from the data obtained during the one or more daytime and nighttime cycles by:
• identifying one or more daytime and nighttime boundaries by:
• selecting from at least one of physical activity, heart rate, or sleep status data;
• temporally segmenting the physical activity, heart rate, or sleep status data to establish a circadian rhythm;
• using at least one of a 2-mean, 5-mean, or K-nearest neighbor algorithm to determine the one or more daytime and nighttime boundaries;
• and
• identifying and predicting the personalized circadian rhythm.