US 12,147,927 B1
Machine learning system and method for predicting caregiver attrition
Jonathan J. Hull, San Carlos, CA (US); and Geoffrey Nudd, San Francisco, CA (US)
Assigned to CLEARCARE, INC., San Francisco, CA (US)
Filed by ClearCare, Inc., San Francisco, CA (US)
Filed on Jul. 29, 2022, as Appl. No. 17/877,834.
Application 17/877,834 is a continuation of application No. 16/102,559, filed on Aug. 13, 2018, granted, now 11,436,549.
Claims priority of provisional application 62/558,342, filed on Sep. 13, 2017.
Claims priority of provisional application 62/545,350, filed on Aug. 14, 2017.
Int. Cl. G06Q 30/00 (2023.01); G06F 16/9535 (2019.01); G06N 20/00 (2019.01); G06Q 10/0635 (2023.01); G06Q 10/0639 (2023.01); H04W 4/38 (2018.01)
CPC G06Q 10/06398 (2013.01) [G06F 16/9535 (2019.01); G06N 20/00 (2019.01); G06Q 10/0635 (2013.01); H04W 4/38 (2018.02)] 19 Claims
OG exemplary drawing
 
1. A method of generating retention information about caregivers responsible for in-home patient care, comprising:
generating patient-caregiver interaction data by analyzing vocal communications occurring in interactions between patients and caregivers during in-home patient care monitored via a voice assistant disposed in a home of a patient having a home assistant configured for each caregiver to recognize speech patterns, participants in conversations, voice stress of each speaker, and words or phrases indicative of stress;
generating caregiver satisfaction data including data associated with feedback caregivers enter from respective mobile devices of caregivers including at least one of caregiver surveys, hours worked, and clock-in/clock out data;
predicting an attrition risk for each caregiver utilizing a machine learning system trained to determine an attrition risk probability for each caregiver based on features of the patient-caregiver interaction data and the caregiver satisfaction data, including features of the vocal communications between patients and caregivers during in-home patient care;
generating, based on the attrition risk, a user interface having graphical elements representing a risk of attrition of individual caregivers providing in-home care of patients;
wherein the machine learning system comprises an ensemble of classifiers to make predictions and the method includes making a population adaptive selection of a classifier, from an ensemble of classifiers, for each caregiver based at least in part on a length of time each caregiver has worked as a caregiver to account for a higher rate of attrition in an initial phase in employment of a caregiver;
wherein each classifier in the ensemble of classifiers is trained to predict a risk of attrition for a different length of employment and the risk of attrition is determined for an individual caregiver by selecting a classifier, from the ensemble of classifiers, corresponding to the length of employment of the individual caregiver, and utilizing the selected classifier to predict the risk of attrition; and
wherein the method comprises utilizing the ensemble of classifiers to predict a risk of attrition of an individual caregiver who has a current length of employment of N weeks will attrit within a next M weeks, where N and M are positive integers.