US 12,374,433 B1
Decision support system using an artificial intelligence (AI) framework for suggesting codes for a clinical record for a patient and related methods and computer program products
Feili Yu, Shoreline, WA (US); Alex Londeree, Austin, TX (US); Lei Qi, Seattle, WA (US); Bo Han, Bothell, WA (US); Wenji Zhang, Redmond, WA (US); and Chenda Deng, Fort Collins, CO (US)
Assigned to CHANGE HEALTHCARE HOLDINGS LLC, Nashville, TN (US)
Filed by Change Healthcare Holdings LLC, Nashville, TN (US)
Filed on Mar. 29, 2022, as Appl. No. 17/656,963.
Int. Cl. G16H 10/60 (2018.01); G06N 3/045 (2023.01); G06N 3/08 (2023.01)
CPC G16H 10/60 (2018.01) [G06N 3/045 (2023.01); G06N 3/08 (2013.01)] 16 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
receiving, by one or more processors, a record containing clinical information associated with a first patient;
determining a subset of the clinical information based on discarding, by the one or more processors executing a multi-stage machine learning model that is trained using historic patient clinical information records, a feature of the clinical information;
generating, by the one or more processors executing the multi-stage machine learning model, a code from among a set of candidate codes for one or more portions of the subset;
wherein generating the code by the multi-stage machine learning model comprises:
detecting, by a first machine-learned model, the subset as being associated with the first patient from among a set of data comprising data associated with a second patient and the first patient;
detecting, by a second machine-learned model and based at least in part on the subset, a set of encounters contained in the subset;
detecting, by a third machine-learned model based at least in part on a first encounter of the set of encounters, a first portion of the clinical information;
generating, by a fourth machine-learned model and based at least in part on the first portion, a first embedding;
retrieving a set of second embeddings generated by the fourth machine-learned model based on respective descriptions of the set of candidate codes;
determining, by the fourth machine-learned model, a set of similarity scores based on a distance between respective ones of the set of second embeddings and the first embedding; and
determining the code to associate with the first portion based on the set of similarity scores;
causing presentation of a partition of the set of encounters, including an indication of the first portion and the code;
receiving user input indicating approval of the code; and
authorizing payment associated with the first encounter determined based at least in part on the approval and the code.