System for supporting decision-making regarding assessment using a machine learning-trained model, assessment support method, and recording medium

Patent metadata

Field	Value
Publication number	US-2026094718-A1
Application number	US-202519414398-A
Country	US
Kind code	A1
Filing date	Dec 10, 2025
Priority date	May 27, 2021
Publication date	Apr 2, 2026
Grant date	—

Abstract

An assessment support system includes: an assessment prediction unit that predicts, based on patient information about a target patient who is a creation target for an assessment in a nursing record, an assessment vector obtained by vectorizing the assessment of the target patient, as a prediction assessment vector; a degree-of-similarity calculation unit that calculates a degree of similarity of the assessment vector to the prediction assessment vector, based on a relationship between the predicted prediction assessment vector and the assessment vector of a patient having the assessment recorded in the nursing record; and a search unit that searches for and outputs at least one similar patient who is similar to the target patient, based on the degree of similarity.

First claim

1 . An assessment support system comprising:
at least one memory that is configured to store instructions; and at least one processor that is configured to execute the instructions to:
generate a first plurality of assessment vectors in a common multi-dimensional assessment space from first patient information in first nursing records of a plurality of first patients by applying a word embedding model to assessment text contained in first nursing records associated with the first patient information;
train a prediction model based on a training set comprising the first plurality of assessment vectors and the first patient information, wherein the prediction model is configured to receive patient information about a target patient as an input and output a predicted assessment vector of the target patient, wherein the training comprises adjusting parameter values of the prediction model to reduce a difference between:
a plurality of predicted assessment vectors output by the prediction model based on the first patient information being input into the prediction model, and
corresponding vectors from among the first plurality of assessment vectors;

input the patient information about the target patient into the trained prediction model, in response to operation information received from a terminal, to obtain the predicted assessment vector of the target patient from the trained prediction model;
calculate a plurality of similarity scores between the predicted assessment vector of the target patient and a second plurality of assessment vectors corresponding to a plurality of second patients using at least one of a cosine similarity or a Minkowski-distance-based metric;
identify one or more patients from among the plurality of second patients having highest similarity scores, by applying a threshold to the plurality of similarity scores; and
retrieve assessment information associated with the identified one or more patients from a database and transmit the retrieved assessment information to the terminal to cause the terminal to display the retrieved assessment information on a screen for creating an assessment of the target patient, wherein the retrieved assessment information comprises at least one of: subjective information reported by the one or more patients, objective information recorded with respect to the one or more patients, or a nursing plan associated with the one or more patients.

2 . The assessment support system according to claim 1 ,
wherein the first patient information comprises at least one of physical information about a body, disease information about a disease or sickness, and past assessment information about the assessment created in the past, with respect to the plurality of first patients.

3 . The assessment support system according to claim 1 ,
wherein the at least one processor is configured to execute the instructions to calculate the plurality of similarity scores such that a similarity score is higher as a degree of coincidence increases between the patient information of the target patient and corresponding patient information of the plurality of second patients.

4 . The assessment support system according to claim 1 ,
wherein the at least one processor is configured to execute the instructions to identify the one or more patients in descending order of similarity scores of the identified one or more patients.

5 . The assessment support system according to claim 1 , wherein the at least one processor is configured to execute the instructions to identify the one or more patients based on similarity scores of the one or more patients being greater than or equal to a predetermined value.

6 . An assessment support method that allows at least one computer to execute:
generating a first plurality of assessment vectors in a common multi-dimensional assessment space from first patient information in first nursing records of a plurality of first patients by applying a word embedding model to assessment text contained in first nursing records associated with the first patient information; training a prediction model based on a training set comprising the first plurality of assessment vectors and the first patient information, wherein the prediction model is configured to receive patient information about a target patient as an input and output a predicted assessment vector of the target patient wherein the training comprises adjusting parameter values of the prediction model to reduce a difference between:
a plurality of predicted assessment vectors output by the prediction model based on the first patient information being input into the prediction model, and
corresponding vectors from among the first plurality of assessment vectors;
inputting the patient information about the target patient into the trained prediction model, in response to operation information received from a terminal, to obtain the predicted assessment vector of the target patient from the trained prediction model; calculating a plurality of similarity scores between the predicted assessment vector of the target patient and a second plurality of assessment vectors corresponding to a plurality of second patients using at least one of a cosine similarity or a Minkowski-distance-based metric; identifying one or more patients from among the plurality of second patients having highest similarity scores, by applying a threshold to the plurality of similarity scores; and retrieving assessment information associated with the identified one or more patients from a database and transmit the retrieved assessment information to the terminal to cause the terminal to display the retrieved assessment information on a screen for creating an assessment of the target patient, wherein the retrieved assessment information comprises at least one of: subjective information reported by the one or more patients, objective information recorded with respect to the one or more patients, or a nursing plan associated with the one or more patients.

7 . The assessment support method according to claim 6 ,
wherein the first patient information comprises at least one of physical information about a body, disease information about a disease or sickness, and past assessment information about the assessment created in the past, with respect to the plurality of first patients.

8 . The assessment support method according to claim 6 ,
calculating the plurality of similarity scores such that score is higher as a degree of coincidence increases between the patient information of the target patient and corresponding patient information of the plurality of second patients.

9 . The assessment support method according to claim 6 ,
identifying the one or more patients in descending order of similarity scores of the identified one or more patients.

10 . The assessment support method according to claim 6 ,
identifying the one or more patients based on similarity scores of the one or more patients being greater than or equal to a predetermined value.

11 . A non-transitory recording medium on which a computer program is recorded, the computer program allowing at least one computer to execute:
generating a first plurality of assessment vectors from first patient information in first nursing records of a plurality of first patients by applying a word embedding model to the first patient information; training a prediction model based on a training set comprising the first plurality of assessment vectors and the first patient information, wherein the prediction model is configured to receive patient information about a target patient as an input and output a predicted assessment vector of the target patient wherein the trainin

Assignees

• Nec Corp

Inventors

• Araki Kenji
• Uno Yutaka
• Yahara Junichi
• Nishihara Kosuke

Classifications

• G16H10/60

  for patient-specific data, e.g. for electronic patient records · CPC title

• G16H50/70

  for mining of medical data, e.g. analysing previous cases of other patients · CPC title

• G16H50/20 Primary

  for computer-aided diagnosis, e.g. based on medical expert systems · CPC title