Program, information processing device, and information processing method

Abstract

To provide programs that output advice regarding patients. [Solution] The program causes a computer to perform a process of obtaining the patient's condition, background, and records of medical professionals regarding the patient, based on sensor data obtained from sensors provided for the patient, and then providing the obtained condition, background, and records to a language model to output advice regarding the patient. Preferably, the sensor is a bed sensor 21, and the sleep state of the patient over a predetermined period is identified based on sensor data over a predetermined period obtained from the bed sensor 21.

Description

Technical Field

[0001] The present disclosure relates to a program, an information processing apparatus, and an information processing method.

Background Art

[0002] In Patent Document 1, an information processing apparatus is disclosed that outputs an abstract text in which editing information is reflected by providing the abstract basic information that is the basis of the abstract text and the editing information that assists in generating the abstract text to a language model.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, the invention according to Patent Document 1 has a problem that it cannot output advice regarding a patient.

[0005] In one aspect, an object is to provide a program or the like that outputs advice regarding a patient.

Means for Solving the Problems

[0006] A program according to one aspect causes a computer to execute a process of obtaining the state of the patient, the background of the patient, and the records of medical staff regarding the patient specified based on sensor data obtained from a sensor provided corresponding to the patient, and giving the obtained state, background, and records to a language model to output advice regarding the patient.

Effects of the Invention

[0007] In one aspect, it becomes possible to output advice regarding a patient. [Brief explanation of the drawing]

[0008] [Figure 1] This is an explanatory diagram showing an overview of the information processing system. [Figure 2] This is a block diagram showing an example server configuration. [Figure 3] This is an explanatory diagram showing an example of a patient database record layout. [Figure 4] This is an explanatory diagram showing an example of a record layout in a record database. [Figure 5] This is an explanatory diagram showing an example of a record layout for a sensor database. [Figure 6] This block shows an example of a smartphone configuration. [Figure 7] This is an explanatory diagram showing an example of a prompt. [Figure 8] This is an explanatory diagram showing an example screen. [Figure 9] This is an explanatory diagram showing an example screen. [Figure 10] This is a flowchart that explains the processing flow of a program. [Figure 11] This is an explanatory diagram showing an example of the record layout of the sensor database according to Embodiment 2. [Figure 12] This is an explanatory diagram showing an example of a prompt. [Figure 13] This is a flowchart that explains the processing flow of a program. [Figure 14] This is a block diagram showing an example server configuration. [Figure 15] This is an explanatory diagram showing an example of a record layout in the stamp information database. [Figure 16] This is an explanatory diagram showing an example of a record layout for a caregiving data database. [Figure 17] This is an explanatory diagram showing an example of a registration screen. [Figure 18] This is an explanatory diagram showing an example of a prompt. [Figure 19] This is a flowchart that explains the processing flow of a program. [Figure 20] It is an explanatory diagram showing an example of the record layout of the record DB. [Figure 21] It is an explanatory diagram showing an example of the record layout of the stamp information DB. [Figure 22] It is an explanatory diagram showing an example of a prompt. [Figure 23] It is a block diagram showing a configuration example of the server. [Figure 24] It is an explanatory diagram showing an example of a prompt. [Figure 25] It is a flowchart for explaining the flow of program processing. [Figure 26] It is a flowchart for explaining the flow of program processing.

Mode for Carrying Out the Invention

[0009] (Embodiment 1) FIG. 1 is an explanatory diagram showing an overview of an information processing system. The information processing system provides records of a patient's condition, patient background, and healthcare workers' records regarding the patient to a language model to obtain advice regarding the patient and notify the obtained advice regarding the patient to the healthcare workers. The information processing system includes an information processing device 10, a sensor I / F (Interface) device 20, and a terminal device 30. The information processing device 10, the sensor I / F device 20, and the terminal device 30 are connected via a network.

[0010] The information processing device 10 is an information processing device that performs processing, storage, and transmission / reception of various information. The information processing device 10 is, for example, a server device, a personal computer, or a general-purpose tablet PC (personal computer), etc. Note that the information processing device 10 may be a cloud server device that provides the functions included in the information processing device 10 as a cloud service. In the present embodiment, the information processing device 10 will be described as the server 10.

[0011] The terminal device 30 is, for example, a smartphone, tablet, personal computer, or general-purpose tablet PC (personal computer). In this embodiment, the terminal device 30 will be described as a smartphone 30. Healthcare workers carry the smartphone 30 with them when performing their duties.

[0012] A sensor interface device 20 is provided for each patient. The sensor interface device 20 transmits sensor data obtained from sensors provided for each patient to the server 10. Sensors provided for each patient include bed sensors, excretion sensors, pillow sensors, clip sensors, touch sensors, infrared sensors, mat sensors, camera sensors, or sensors attached to the patient's wheelchair. A sensor attached to the patient's wheelchair detects, for example, drowsiness and napping of the patient sitting in the wheelchair. In this embodiment, an example using a bed sensor as a sensor provided for each patient is described. In Figure 1, the sensor interface device 20 and the bed sensor 21 are connected by wire or wireless. Also in Figure 1, the patients (Patient A and Patient B), the sensor interface device 20, and the bed sensor 21 are associated via patient identification information (e.g., patient ID).

[0013] One bed sensor 21 is installed on each patient's bed. The bed sensor 21 is, for example, a sheet-shaped sensor and is installed between the bed mattress and the sheet. The bed sensor 21 acquires sensor data from the patient lying in bed. The sensor data includes the date and time the sensor data was acquired, heart rate, respiratory rate, and body movement. The heart rate is, for example, the average heart rate per minute. The respiratory rate is, for example, the average number of respiratory rate per minute. The body movement is, for example, the number of times a large vibration exceeding a predetermined threshold (e.g., turning over in bed) is detected per minute. The bed sensor 21 outputs the acquired sensor data to the sensor I / F device 20. The sensor I / F device 20 transmits the sensor data to the server 10 at predetermined time intervals (e.g., every minute).

[0014] Figure 2 is a block diagram showing an example of a server configuration. Server 10 includes a control unit 11, a storage unit 12, a communication unit 13, a large-capacity storage unit 14, and a read unit 15. The above-mentioned units are interconnected via a bus. The control unit 11 is configured using one or more processors, such as a CPU (Central Processing Unit), an MPU (Micro-Processing Unit), or a GPU (Graphics Processing Unit).

[0015] The storage unit 12 includes RAM (Random Access Memory) or ROM (Read Only Memory), etc. The storage unit 12 stores various data necessary for the control program 12P (program product) executed by the control unit 11. The storage unit 12 also temporarily stores data generated when the control program 12P is executed. The control unit 11 performs various information processing and control processing related to the server 10 by appropriately executing the control program 12P stored in the storage unit 12.

[0016] Furthermore, the memory unit 12 stores the language model 12M. The language model 12M is a general-purpose large language model (LLM: Large Language Model) constructed by performing unsupervised pre-training using, for example, a large set of texts and images. The language model 12M can be constructed using algorithms such as GPT (Generative Pre-trained Transformer)-3, GPT-3.5, GPT-4, RWKV (Receptance Weighted Key Value), PaLM2, or LLaMa (Large Language Model Meta AI).

[0017] When a prompt is input that includes the patient's condition, the patient's background, the healthcare worker's record of the patient, and a command to generate advice about the patient, the language model 12M outputs advice about the patient in accordance with the prompt. The patient's condition includes, for example, the patient's sleep state or the patient's excretory state. In Embodiment 1, the patient's condition is described as the patient's sleep state. The patient's sleep state includes, for example, nocturnal awakenings, REM sleep, non-REM sleep, or naps. The patient's background includes the patient's name, sex, age, or symptoms. The healthcare worker's record of the patient includes the date and time and the healthcare worker's record at that date and time.

[0018] The language model 12M may be configured by combining multiple algorithms. Instead of storing the language model 12M in the memory unit 12, the control unit 11 may access a language processing server that stores the language model 12M and input prompts to the language model 12M.

[0019] The communication unit 13 is a communication module that transmits and receives information between the sensor I / F device 20 and the smartphone 30 via the network N. The large-capacity storage unit 14 includes RAM or ROM, etc. The large-capacity storage unit 14 stores the patient DB 141, the record DB 142, and the sensor DB 143. The patient DB 141, the record DB 142, and the sensor DB 143 will be described later.

[0020] In this embodiment, the storage unit 12 and the large-capacity storage unit 14 may be configured as a single storage device. The large-capacity storage unit 14 may be composed of multiple storage devices. The large-capacity storage unit 14 may also be an external storage device connected to the server 10.

[0021] The reading unit 15 reads information stored in the portable storage medium 1a. The portable storage medium 1a is, for example, a CD (Compact Disc), DVD (Digital Versatile Disc), USB (Universal Serial Bus) memory, or SD (Secure Digital). The reading unit 15 reads the control program 12P from the portable storage medium 1a. The control unit 11 stores the read control program 12P in the storage unit 12. The control unit 11 may also download the control program 12P from another computer via the network N. In that case, the control unit 11 stores the downloaded control program 12P in the storage unit 12. The control unit 11 may also store the read control program 12P in the large-capacity storage unit 14.

[0022] In this embodiment, server 10 may be composed of multiple servers. Server 10 may be a virtual machine virtually constructed by software within a single device. Server 10 may be a local server installed in the facility where server 10 is located. Server 10 may be a cloud server connected via network N. Furthermore, the control program 12P may run on a single server, or it may be distributed and run on multiple servers interconnected via network N.

[0023] Figure 3 is an explanatory diagram showing an example of the record layout of the patient database. The patient database 141 includes columns for patient ID, name, gender, age, and medical condition. The patient ID column stores the patient ID to identify the patient. The name column stores the patient's name. The gender column stores the patient's gender. The age column stores the patient's age. The medical condition column includes columns for care level and dementia independence level, etc. The care level column stores the patient's care level. The care level is divided into five stages from 1 to 5. "1" is the lowest care level, and "5" is the highest care level.

[0024] The dementia independence level column stores the level of independence in daily living for patients with dementia. The level of independence in daily living for patients with dementia is divided into nine levels, for example, "- (not dementia)", "I", "II", "IIa", "IIb", "III", "IIIa", "IIIb", and "M". If the patient has dementia, one of "I", "II", "IIa", "IIb", "III", "IIIa", "IIIb", or "M" is stored; if the patient does not have dementia, "- (not dementia)" is stored. In Figure 3, the record for patient ID "K001" stores the name "A", gender "female", age "73", care level "1", and dementia independence level "II".

[0025] Figure 4 is an explanatory diagram showing an example of the record layout of the record database. The record database 142 stores records of healthcare workers for each patient. Records of healthcare workers for each patient are entered via an input screen (not shown) displayed on a smartphone 30. The record database 142 includes multiple record sheets 142a. Multiple record sheets 142a are provided for each patient. In Figure 4, the record sheet 142a for patient A is displayed. The record sheet 142a includes a patient data field R1, a date column, and a record column. The patient data field R1 stores the patient ID. In Figure 4, the patient ID "K001" is stored in the patient data field R1. The date column stores the date on which the healthcare worker's record for the patient was received. The record column stores the healthcare worker's record for the patient. In Figure 4, the record for the date "2024 / 4 / 29" stores the healthcare worker's record for the patient, "Complaint of poor health." In this embodiment, an example of a medical professional's record of a patient is described using keywords such as "complaint of feeling unwell," but it is not limited to this. A medical professional's record of a patient may also be, for example, a sentence such as "Patient A appears depressed," or an image of the patient's affected area.

[0026] Figure 5 is an explanatory diagram showing an example of the record layout of the sensor DB. The sensor DB 143 stores sensor data for each patient. The sensor DB 143 includes multiple sensor sheets 143a. Multiple sensor sheets 143a are provided for each patient. In Figure 5, the sensor sheet 143a for patient A is displayed. The sensor sheet 143a includes a patient data field R1, a date and time column, a respiratory rate column, a heart rate column, and a body movement volume column. The patient ID is stored in the patient data field R1. In Figure 5, the patient ID "K001" is stored in the patient data field R1. The date and time column stores the date and time when sensor data was received from the sensor I / F device 20. The respiratory rate column stores the patient's respiratory rate. The heart rate column stores the patient's heart rate. The body movement volume column stores the patient's body movement volume. The record for the date and time "2024 / 4 / 10 / 7:00" in Figure 5 contains the following data: respiratory rate "56 (breaths / min)", heart rate "20 (beats / min)", and body movement "4 (beats / min)".

[0027] Figure 6 is a block diagram showing an example of a smartphone configuration. The smartphone 30 includes a control unit 31, a storage unit 32, a communication unit 33, an input unit 34, and a display unit 35. Each of the above-mentioned units is interconnected via a bus. The control unit 31 is configured using one or more processors such as a CPU, MPU, or GPU. The storage unit 32 includes RAM or ROM. The storage unit 32 stores various data necessary for the control program 32P (program product) executed by the control unit 31. The storage unit 32 temporarily stores data generated when the control program 32P is executed. The control unit 31 performs information processing and control processing related to the smartphone 30 by appropriately executing the control program 32P stored in the storage unit 32.

[0028] The communication unit 33 is a communication module that sends and receives information to and from the server 10 via the network N. The input unit 34 accepts operation input from medical personnel. The input unit 34 is, for example, a microphone for voice input. The display unit 35 is a liquid crystal display or an organic EL display, etc. The display unit 35 outputs various information according to instructions from the control unit 31. The input unit 34 and the display unit 35 may be integrated into a single touch panel.

[0029] In the following description, we will use the case where a medical professional uses a smartphone 30 with the program of this embodiment already installed as an example. When the control unit 31 executes the program of this embodiment, it displays an initial screen (not shown) on the display unit 35. When the initial screen is displayed, the control unit 31 may perform authentication processing for the medical professional. In that case, the authentication processing for the medical professional may use, for example, authentication by the medical professional's identification information (e.g., name or ID), facial recognition, or fingerprint recognition. The medical professional inputs the patient's identification information via the input unit 34. The patient's identification information may be, for example, the patient ID or the patient's name. In this embodiment, the patient's identification information will be described as the patient ID. The control unit 31 obtains the patient ID entered by the medical professional. In the following description, we will describe an example where the control unit 31 obtains the patient ID "K001". The control unit 31 sends the obtained patient ID "K001" and a notification that the program has been started (hereinafter referred to as the start notification) to the server 10.

[0030] The control unit 11 receives the patient ID "K001" and activation notification transmitted from the smartphone 30. The control unit 11 reads the patient background information (name "A", gender "female", age "75", care level "1", and dementia independence level "II") corresponding to patient ID "K001" from the patient database 141. The control unit 11 reads the records of healthcare workers for the patient associated with patient ID "K001" ("2024 / 4 / 9 Complaint of feeling unwell", "2024 / 1 / 26 Discharged", and "2024 / 1 / 7 Fracture due to fall") from the record database 142.

[0031] The control unit 11 reads sensor data for a predetermined period corresponding to patient ID "K001" from the sensor DB 143. The predetermined period includes, for example, one day ago, the past week, or the past month. In this embodiment, the predetermined period will be described as "the past week". In other words, if the control unit 11 receives a startup notification at "2024 / 4 / 10 / 9:15", it reads sensor data from the sensor DB 143 from the date and time "2024 / 4 / 10 / 7:00" to the date and time "2024 / 4 / 3 / 0:00".

[0032] The control unit 11 identifies the patient's sleep state over the past week based on the sensor data read from the past week. The control unit 11 also identifies the sleep state, including the type of sleep and the number of occurrences of each type of sleep during the first period, and the type of sleep and the number of occurrences of each type of sleep during the second period, which includes the first period, based on the sensor data read from the past week.

[0033] First, we will explain how to identify the sleep state over the past week from sensor data from the past week. The sleep state over the past week may include, for example, the duration of nighttime awakenings, the duration of REM sleep, the duration of non-REM sleep, the types of sleep and the number of occurrences of each type of sleep, or the types of sleep and the duration of each type of sleep. In this embodiment, the sleep state over the past week will be described as "the duration of nighttime awakenings over the past week." The control unit 11 extracts each sensor data that constitutes the sensor data for the past week (for example, date and time "2024 / 4 / 10 / 7:00", respiratory rate "56 (breaths / min)", heart rate "20 (beats / min)", and body movement "4 (times / min)").

[0034] The control unit 11 identifies four types of sleep (e.g., middle-of-the-night awakening, REM sleep, non-REM sleep, or nap) from the extracted sensor data using a rule-based or machine learning model. When using a machine learning model to identify the four types of sleep from the sensor data, the machine learning model takes the date and time, respiratory rate, heart rate, and body movement as input and outputs the four types of sleep. The machine learning model is generated from a large amount of training data in which input data including date and time, respiratory rate, heart rate, and body movement are associated with the four types of sleep. The control unit 11 inputs the date and time, respiratory rate, heart rate, and body movement of each sensor data into the machine learning model. The control unit 11 identifies the sleep type directly from the output result of the machine learning model (e.g., middle-of-the-night awakening, REM sleep, non-REM sleep, or nap).

[0035] The machine learning model may take respiratory rate, heart rate, and body movement as input and output one of three sleep types (e.g., nocturnal awakening, REM sleep, or non-REM sleep). In this case, the machine learning model is generated from a large amount of training data, which associates input data including respiratory rate, heart rate, and body movement with the three sleep types. The control unit 11 inputs the respiratory rate, heart rate, and body movement of each sensor data into the machine learning model. If the output result of the machine learning model is "nocturnal awakening," the control unit 11 identifies the sleep type as "nocturnal awakening." If the output result of the machine learning model is "REM sleep" or "non-REM sleep," the control unit 11 determines whether the date and time of each sensor data falls within the daytime period (e.g., 11:00-16:00). If the date and time of each sensor data falls within the daytime period, the control unit 11 identifies the sleep type as "afternoon nap." If the date and time of each sensor data does not fall within the daytime period, the control unit 11 identifies the sleep type as "REM sleep" or "non-REM sleep."

[0036] The control unit 11 identifies the type of sleep for all sensor data that make up the sensor data for the past week, using a rule-based or machine learning model. The control unit 11 then aggregates the times of sensor data in which the type of sleep was identified as waking up in the middle of the night, thereby identifying the time when waking up in the middle of the night occurred during the past week.

[0037] Next, a method for identifying the sleep state, including the type of sleep and the number of occurrences of each type of sleep during the patient's first period, and the type of sleep and the number of occurrences of each type during the second period including the first period, will be described from the sensor data read out over the past week. The first period includes, for example, the day before, the past three days, the past week, or the past month. The second period is a past period including the first period, and includes, for example, the past three days, the past week, or the past month. In this embodiment, the first period will be described as "the past three days," and the second period will be described as "the past week."

[0038] The control unit 11 extracts each sensor data that makes up the sensor data for the past week. The control unit 11 identifies the type of sleep for each extracted sensor data using a rule-based or machine learning model. The control unit 11 identifies the type of sleep for all sensor data that makes up the sensor data for the past week using a rule-based or machine learning model.

[0039] The control unit 11 identifies the patient's sleep state (hereinafter referred to as the "sleep state of the first period"), including the type of sleep and the number of occurrences of each type over the past three days, by aggregating the sensor data for the past three days from the sensor data for the past week, categorized by sleep type. The control unit 11 identifies the patient's sleep state (hereinafter referred to as the "sleep state of the second period"), including the type of sleep and the number of occurrences of each type over the past week, by aggregating the sensor data for the past week, categorized by sleep type. The following describes an example using the sleep states of the first and second periods as the patient's sleep state.

[0040] The control unit 11 generates a prompt 40 (hereinafter referred to as "prompt 40") which includes the sleep state of the identified patient during the first and second periods, the patient's background, records of medical professionals regarding the patient, and a command to generate advice about the patient. The command to generate advice about the patient is pre-stored in the memory unit 12 or the large-capacity memory unit 14. The command to generate advice about the patient can be modified as appropriate according to the embodiment. Figure 7 is an explanatory diagram showing an example of a prompt. Prompt 40 includes a background field 41, a sleep state field 42, a record field 43, and a generation command field 44. The background field 41 contains the patient's background information read from the patient DB 141. In Figure 7, the background field 41 contains the following: "The patient's name is Ms. A. Ms. A is a 75-year-old woman with care level 1 and dementia level 2."

[0041] The sleep status column 42 contains the patient's sleep status, as identified based on sensor data obtained from the bed sensor 21. In Figure 7, the sleep status column 42 states: "Patient A's recent sleep status is as follows: There were 5 instances of awakening during the night in the past 3 days. Additionally, there were 13 instances of awakening during the night in the past week. There were 3 instances of napping in the past 3 days. Additionally, there were 6 instances of napping in the past week."

[0042] Record field 43 contains the medical staff's records for the patient, read from the record DB 142. In Figure 7, record field 43 contains the following: "The medical staff's records for Mr. / Ms. A are as follows: On April 9, 2024, Mr. / Ms. A complained of feeling unwell. On January 26, 2024, Mr. / Ms. A fractured bone due to a fall on January 7, 2024..." In this embodiment, an example has been described in which all the information recorded in the record DB 142 is recorded in record field 43, but this is not the only example. Record field 43 may also contain information stored over the past three days or the past week. If information stored over the past three days is recorded, record field 43 will contain the following: "The medical staff's records for Mr. / Ms. A complained of feeling unwell on April 9, 2024."

[0043] The generation command field 44 contains instructions for the language model 12M to output advice to the patient. In Figure 7, the generation command field 44 states: "As a medical and nursing specialist, you should generate advice to be conveyed to the medical professionals responsible for A's care. When generating the advice, consider the diseases and illnesses that A should be aware of based on the current situation, points that A should be aware of in daily life, and points that medical professionals should be aware of today."

[0044] The control unit 11 inputs the generated prompt 40 to the language model 12M, thereby obtaining patient advice generated by the language model 12M. The control unit 11 then sends the obtained patient advice to the smartphone 30.

[0045] The control unit 31 receives patient advice transmitted from the server 10. The control unit 31 displays the received patient advice on the display unit 35. Figure 8 is an explanatory diagram showing an example screen. Screen d1 in Figure 8 includes area d11. Patient advice is displayed in area d11, item by item.

[0046] (modified version) Embodiment 1 describes an example in which advice about the patient is displayed on the healthcare worker's smartphone 30, but it is not limited to this. In addition to advice about the patient, the patient's sleep status and the healthcare worker's records regarding the patient may also be displayed on the smartphone 30. The process in that case will be described below. The control unit 11 inputs the generated prompt 40 to the language model 12M, thereby obtaining the advice about the patient generated by the language model 12M. The control unit 11 transmits the advice about the patient, the sleep status for the first and second periods, and the healthcare worker's records regarding the patient to the smartphone 30.

[0047] The control unit 31 receives patient advice, sleep status for the first and second periods, and records of medical professionals' interactions with the patient, transmitted from the server 10. The control unit 31 displays the received patient advice, sleep status for the first and second periods, and records of medical professionals' interactions with the patient on the display unit 35. Figure 9 is an explanatory diagram showing an example screen. Screen d2 in Figure 9 includes areas d11, d12, and d13. The format of area d11 is the same as in Embodiment 1. Area d12 displays the sleep status for the first and second periods. Area d13 displays records of medical professionals' interactions with the patient.

[0048] Figure 10 is a flowchart illustrating the program's processing flow. When the program of this embodiment is executed, the control unit 31 displays an initial screen (not shown) on the display unit 35 (step S101). The control unit 31 obtains the patient ID entered by the medical professional (step S102). The control unit 31 sends the obtained patient ID and a startup notification to the server 10 (step S103).

[0049] The control unit 11 receives the patient ID and activation notification transmitted from the smartphone 30 (step S201). The control unit 11 reads the patient background corresponding to the patient ID from the patient DB 141 (step S202). The control unit 11 reads the records of healthcare workers for the patient associated with the patient ID from the record DB 142 (step S203). The control unit 11 reads the sensor data for the past week corresponding to the patient ID from the sensor DB 143 (step S204).

[0050] The control unit 11 aggregates the sensor data read from the past week to identify the sleep state for the first and second periods (step S205). The control unit 11 generates a prompt 40 by combining the sleep state for the first and second periods, the patient's background, the medical professional's records regarding the patient, and a command to generate advice regarding the patient (step S206). The control unit 11 inputs the generated prompt 40 to the language model 12M to obtain the patient advice generated by the language model 12M (step S207). The control unit 11 transmits the obtained patient advice, the sleep state for the first and second periods, and the medical professional's records regarding the patient to the smartphone 30 (step S208). The control unit 11 may transmit only the obtained patient advice to the smartphone 30.

[0051] The control unit 31 receives patient advice, sleep status during the first and second periods, and records of medical professionals regarding the patient, transmitted from the smartphone 30 (step S104). The control unit 31 displays the received patient advice, sleep status during the first and second periods, and records of medical professionals regarding the patient on the display unit 35 (step S105). The control unit 31 may also display only the received patient advice on the display unit 35.

[0052] According to Embodiment 1, the information processing system can output advice about the patient by providing the patient's sleep state, the patient's background, and records of healthcare professionals' interactions with the patient to a language model.

[0053] According to Embodiment 1, the information processing system can identify the patient's sleep state over a predetermined period based on sensor data obtained from a bed sensor over a predetermined period.

[0054] According to Embodiment 1, the information processing system can identify the sleep state, including the type of sleep and the number of occurrences of each type during the patient's first period, and the type of sleep and the number of occurrences of each type during the second period, which includes the first period, based on sensor data for a predetermined period obtained from the bed sensor.

[0055] According to Embodiment 1, the information processing system can output advice regarding the patient by providing a language model with the sleep state during the first and second periods, the patient's background, and records of the medical professionals' interactions with the patient.

[0056] According to Embodiment 1, the information processing system can display patient-related advice on the healthcare professional's smartphone.

[0057] In Embodiment 1, the information processing system can display, on the healthcare worker's smartphone, the patient's sleep status and the healthcare worker's records regarding the patient, in addition to providing advice about the patient.

[0058] (Embodiment 2) Embodiment 2 describes an example in which an information processing system provides a language model with the patient's sleep state, the patient's background, and records of medical professionals' interactions with the patient, as well as the patient's excretion status over the past week (a predetermined period) identified based on sensor data from the past week (a predetermined period).

[0059] The bed sensor 21 of Embodiment 2 detects whether or not the patient has excreted (defecated or urinated). The bed sensor 21 detects whether or not the patient has excreted by measuring the change in electrical resistance value when there is no excretion such as feces and urine on the surface of the bed sensor 21 and when there is excretion such as feces and urine on the surface of the bed sensor 21. Alternatively, the bed sensor 21 may detect whether or not the patient has excreted by measuring the decrease in the resonant frequency when there is no excretion such as feces and urine on the surface of the bed sensor 21 and when there is excretion such as feces and urine on the surface of the bed sensor 21.

[0060] When the bed sensor 21 detects patient excretion, it outputs a notification (hereinafter referred to as "excretion notification") to the sensor I / F device 20 indicating that patient excretion has been detected. When the bed sensor 21 detects patient excretion, the sensor I / F device 20 transmits the sensor data, including the excretion notification, to the server 10.

[0061] The information processing system may use a diaper sensor placed on the patient to detect patient excretion instead of the bed sensor 21. Alternatively, the information processing system may detect patient excretion based on reports from the patient or healthcare worker. Furthermore, the information processing system may detect patient excretion based on healthcare worker records or stamp information (see Embodiment 3).

[0062] Figure 11 is an explanatory diagram showing an example of the record layout of the sensor DB according to Embodiment 2. The format of the sensor DB 143 is the same as in Embodiment 1. The sensor sheet 143a includes an excretion column. The excretion column stores whether or not the patient has excreted. When an excretion notification is received from the sensor I / F device 20, "〇" is stored in the excretion column, and when no excretion notification is received from the sensor I / F device 20, "―" is stored in the excretion column. In the record for the date and time "2024 / 4 / 10 / 7:00" in Figure 11, the respiratory rate is "56 (breaths / min)", heart rate is "20 (beats / min)", body movement is "4 (times / min)", and excretion is "〇" is stored.

[0063] The processing of Embodiment 2 will now be described. In the following description, an example will be given in which the control unit 31 acquires the patient ID "K001". The control unit 31 sends the acquired patient ID "K001" and a startup notification to the server 10.

[0064] The control unit 11 receives the patient ID "K001" and activation notification transmitted from the smartphone 30. The control unit 11 reads the patient background corresponding to patient ID "K001" from the patient DB 141. The control unit 11 reads the records of healthcare workers for the patient associated with patient ID "K001" from the record DB 142. The control unit 11 reads the sensor data for the past week corresponding to patient ID "K001" from the sensor DB 143. Based on the sensor data for the past week that has been read, the control unit 11 identifies the sleep state of the patient during the first and second periods.

[0065] Furthermore, the control unit 11 identifies the patient's excretion status over the past week based on the sensor data read from the past week. The method for identifying the excretion status over the past week from the sensor data over the past week will be described below. The excretion status over the past week is, for example, the number of times the patient excreted in bed during the past week, or the duration of bed excretion during the past week. In this embodiment, the excretion status over the past week will be described as "the number of times the patient excreted in bed during the past week." The control unit 11 extracts records from the sensor data over the past week in which the excretion column contains a "〇". The control unit 11 identifies the number of times the patient excreted in bed during the past week by aggregating the number of extracted records.

[0066] The method for identifying the excretion status over the past week (a predetermined period) from sensor data over the past week (a predetermined period) has been described, but it is not limited to this. The control unit 11 may also identify the excretion status over the past three days (a first period) or the excretion status over the past week (a second period) from sensor data over the past week (a predetermined period). In that case, the prompts input to the language model 12M will include the number of times bed excretion occurred over the past three days and the number of times bed excretion occurred over the past week. Specifically, the prompts will include phrases such as, "Bed excretion was identified twice in the past three days. Bed excretion was identified four times in the past week."

[0067] The control unit 11 generates a prompt 50 (hereinafter referred to as "prompt 50") which includes the number of times bedtime excretion occurred in the past week, the patient's sleep status during the first and second periods, the patient's background, records of healthcare professionals regarding the patient, and a command to generate advice regarding the patient.

[0068] Figure 12 is an explanatory diagram showing an example of a prompt. Prompt 50 includes a background field 41, an excretion status field 51, a sleep status field 42, a record field 43, and a generation command field 44. The contents described in the background field 41, sleep status field 42, record field 43, and generation command field 44 are the same as in Embodiment 1. The excretion status field 51 contains the number of times bed excretion occurred in the past week, as identified by the method described above. In Figure 12, the excretion status field 51 states, "Mr. A's recent excretion status is as follows: Bed excretion was identified 4 times in the past week."

[0069] The control unit 11 inputs the generated prompt 50 to the language model 12M, thereby obtaining the patient-related advice generated by the language model 12M. The subsequent processing is the same as in Embodiment 1, so the explanation is omitted.

[0070] Figure 13 is a flowchart illustrating the program's processing flow. The flowchart in Figure 13 shows the same process as in Figure 10, but with steps S204-S207 changed to steps S301-S305. Steps identical to those in Figure 10 are omitted from the explanation.

[0071] The control unit 11 reads sensor data for the past week corresponding to the patient ID from the sensor DB 143 (step S301). The control unit 11 identifies the sleep state for the first and second periods by aggregating the sensor data for the past week that has been read (step S302). The control unit 11 identifies the number of times bedtime excretion occurred in the past week by aggregating the number of records in the excretion column that are stored among the sensor data for the past week that has been read (step S303). The control unit 11 generates a prompt by combining the sleep state for the first and second periods, the number of times bedtime excretion occurred in the past week, the patient's background, the medical professional's record regarding the patient, and the command to generate advice about the patient (step S304). The control unit 11 inputs the generated prompt to the language model 12M and obtains the patient advice generated by the language model 12M (step S305). The control unit 11 proceeds to step S208.

[0072] According to Embodiment 2, the information processing system can identify the number of times excretion has occurred in the past week based on sensor data obtained from the bed sensor over the past week.

[0073] In Embodiment 2, the information processing system can output advice about the patient by providing a language model with sleep status during the first and second periods, the number of times excretion occurred in the past week, the patient's background, and records of healthcare professionals regarding the patient.

[0074] (Embodiment 3) Embodiment 3 describes an example in which an information processing system provides a language model with information on the patient's sleep state, the patient's background, and records of medical professionals' care for the patient, as well as the details of the care provided by medical professionals to the patient.

[0075] Figure 14 is a block diagram showing an example of the server configuration. The large-capacity storage unit 14 stores the stamp information DB 144 and the care content DB 145. The stamp information DB 144 stores stamp information for each patient. The stamp information is, for example, an illustration or image showing the care provided by a medical professional to the patient. The care content DB 145 stores the care provided by a medical professional corresponding to the stamp information.

[0076] Figure 15 is an explanatory diagram showing an example of the record layout of the stamp information database. The stamp information database 144 contains multiple stamp information sheets 144a. Multiple stamp information sheets 144a are provided for each patient. In Figure 15, the stamp information sheet 144a for patient A is displayed. The stamp information sheet 144a includes a patient data field R1, a date and time column, and a stamp information column. The patient data field R1 stores the patient ID. In Figure 15, the patient ID "K001" is stored in the patient data field R1. The date and time column stores the date and time when the selection of stamp information by the medical professional was received. The stamp information column stores the stamp information selected by the medical professional. In Figure 15, the record for the date and time "2024 / 4 / 7 / 22:30" stores stamp information indicating a change in patient position.

[0077] Figure 16 is an explanatory diagram showing an example of a record layout in the care content database. The care content database 145 includes a stamp information column and a care content column. The stamp information column stores stamp information. The care content column stores the care content provided by the medical professional corresponding to the stamp information. In the top record of Figure 16, stamp information indicating a change in body position and the change in body position are stored in association.

[0078] The method for creating the stamp information DB144 will be explained. Based on the instructions of the medical professional, the control unit 31 displays the stamp information registration screen on the display unit 35. Figure 17 is an explanatory diagram showing an example of the registration screen. The registration screen d3 is displayed on the display unit 35 of the smartphone 30. The control unit 11 accepts various inputs for the registration screen d3 via the input unit 34. The registration screen d3 has, from top to bottom, a patient ID field d31, a date and time field d32, a stamp information field d33, and a registration button b1. The patient ID is entered in the patient ID field d31 by the medical professional. In Figure 17, "K001" is entered in the patient ID field d31. The patient's name may also be entered in the patient ID field d31.

[0079] The date and time field d32 is a pull-down menu button. The date and time field d32 displays the date and time when the medical professional selected the stamp information. In Figure 17, "2024 / 4 / 7 / 22:30" is selected in the date and time field d32. The date and time may also be directly entered by the medical professional in the date and time field d32. Multiple stamp information fields d33 are displayed for selection. The stamp information field d33 accepts the selection of target stamp information from multiple stamp information fields. In Figure 17, the stamp information field d33 shows that stamp information indicating a change in patient position is selected. Multiple stamp information may be selected. The registration button b1 is a button for sending the patient ID, date and time, and stamp information entered or selected on the registration screen d3 to the server 10.

[0080] The control unit 31 determines whether or not the selection of the registration button b1 has been accepted. If the control unit 31 determines that the selection of the registration button b1 has not been accepted, it again accepts input or selection in the patient ID field d31, the date and time field d32, and the stamp information field d33. If the control unit 31 determines that the selection of the registration button b1 has been accepted, it sends the input or selected patient ID, date and time, and stamp information to the server 10 on the registration screen d3. The control unit 11 receives the patient ID, date and time, and stamp information sent from the smartphone 30. The control unit 11 stores the received date and time and stamp information in the stamp information DB 144 for each patient ID. The information processing system creates the stamp information DB 144 by repeating the above process.

[0081] The processing of Embodiment 3 will now be described. In the following description, an example will be given in which the control unit 31 acquires the patient ID "K001". The control unit 31 sends the acquired patient ID "K001" and a startup notification to the server 10.

[0082] The control unit 11 receives the patient ID "K001" and activation notification transmitted from the smartphone 30. The control unit 11 reads the patient background corresponding to patient ID "K001" from the patient DB 141. The control unit 11 reads the records of healthcare workers for the patient associated with patient ID "K001" from the record DB 142. The control unit 11 reads the sensor data for the past week corresponding to patient ID "K001" from the sensor DB 143. Based on the sensor data for the past week that has been read, the control unit 11 identifies the sleep state of the patient during the first and second periods.

[0083] The control unit 11 reads stamp information corresponding to patient ID "K001" from the stamp information DB 144. The control unit 11 obtains the care details provided by medical professionals to the patient associated with the read stamp information from the care details DB 145. In Figures 15 and 16, the control unit 11 obtains "position change," "rehabilitation," and "oral care" from the care details DB 145 as care details provided by medical professionals to the patient associated with the stamp information.

[0084] The control unit 11 generates a prompt 60 (hereinafter referred to as "prompt 60") which includes the acquired care details of the medical professional for the patient, the patient's sleep status during the first and second periods, the patient's background, the medical professional's records regarding the patient, and a command to generate advice regarding the patient. Figure 18 is an explanatory diagram showing an example of a prompt. The prompt 60 includes a background field 41, a care details field 61, a sleep status field 42, a record field 43, and a generation command field 44. The contents described in the background field 41, the sleep status field 42, the record field 43, and the generation command field 44 are the same as in Embodiment 1.

[0085] The care details section 61 contains the care details corresponding to the stamp information selected by the medical professional. In Figure 18, the care details section 61 states: "The care provided by the medical professional to Mr. A is as follows: Position change was performed at 22:30 on April 7, 2024. Rehabilitation was performed at 8:30 on April 5, 2024. Oral care was performed at 12:30 on February 5, 2024..."

[0086] In this embodiment, an example has been described in which the care details field 61 contains care details corresponding to all the stamp information recorded in the stamp information DB 144, but it is not limited to this. The care details field 61 may also contain care details corresponding to stamp information stored in the past three days or the past week. If information stored in the past three days is recorded, the record field 43 will contain the following: "The care details provided by medical professionals to Mr. A are as follows: A change of position was performed on April 7, 2024."

[0087] The control unit 11 inputs the generated prompt 60 to the language model 12M, thereby obtaining the patient-related advice generated by the language model 12M. The subsequent processing is the same as in Embodiment 1, so the explanation is omitted.

[0088] Figure 19 is a flowchart illustrating the program's processing flow. The flowchart in Figure 19 shows the same process as in Figure 10, but with steps S206-S207 changed to steps S401-S404. Steps similar to those in Figure 10 are omitted from explanation.

[0089] The control unit 11 reads stamp information corresponding to the patient ID from the stamp information DB 144 (step S401). The control unit 11 obtains the care details of the medical professionals for the patient related to the read stamp information from the care details DB 145 (step S402). The control unit 11 generates a prompt 60 by combining the obtained care details, the patient's sleep state during the first and second periods, the patient's background, the medical professionals' records for the patient, and a command to generate advice about the patient (step S403). The control unit 11 inputs the generated prompt 60 to the language model 12M and obtains the patient advice generated by the language model 12M (step S404). The control unit 11 proceeds to step S208.

[0090] (modified version) In Embodiments 1 to 3, an example was described in which the control unit 11 identifies the patient's sleep state based on sensor data obtained from the bed sensor 21 and includes the identified sleep state in the prompt, but the system is not limited to this. The control unit 11 may also include in the prompt the patient's sleep record extracted from the medical professional's records of the patient and the patient's sleep record identified from the stamp information. The processing in that case will be described below. The patient's sleep record is part of the medical professional's records of the patient, for example, a record of the patient's sleep other than in bed.

[0091] Figure 20 is an explanatory diagram showing an example of the record layout of the record database. The format of record database 142 is the same as in Figure 4. Record database 142 stores the patient's sleep records. The record for the date "2024 / 4 / 10" in Figure 20 stores the patient's sleep record: "The patient complained of insomnia and watched television on a chair in the dining room from around 8 PM. After watching television, the patient slept on the chair for about 3 hours. After that, the patient returned to their room and slept."

[0092] Figure 21 is an explanatory diagram showing an example of the record layout of the stamp information DB. The format of the stamp information DB 144 is the same as in Figure 15. The date and time column stores the date and time, as well as the date and time when the patient's sleep record was obtained. Records in Figure 21 that contain "2024 / 4 / 9 / 13:30~14:45" in the date and time column store stamp information indicating a nap (in a chair, etc.). Records in Figure 21 that contain "2024 / 4 / 8 / 12:08~14:05" in the date and time column store stamp information indicating a nap (in a wheelchair).

[0093] The care details database 145 stores patient sleep records (napping in a chair, etc., and napping in a wheelchair) corresponding to the two types of napping stamp information mentioned above. In addition, if the two types of napping stamp information are selected in the stamp information field d33 of the registration screen d3, a pull-down menu button for entering the time period when the napping occurred will be displayed in the date and time field d32.

[0094] The following describes the processing of a modified example. The control unit 31 sends the acquired patient ID "K001" and the activation notification to the server 10. The control unit 11 receives the patient ID "K001" and the activation notification sent from the smartphone 30. The control unit 11 reads the patient background corresponding to patient ID "K001" from the patient DB 141. The control unit 11 reads the records of healthcare workers for the patient associated with patient ID "K001" from the record DB 142.

[0095] The control unit 11 extracts the patient's sleep record from the read records. Specifically, if the read records contain keywords related to sleep such as "sleep," "going to bed," "sleeping," and "napping," the control unit 11 extracts those records as the patient's sleep record. The control unit 11 reads the sensor data for the past week corresponding to patient ID "K001" from the sensor DB 143. Based on the read sensor data for the past week, the control unit 11 identifies the patient's sleep state during the first and second periods.

[0096] The control unit 11 reads stamp information corresponding to patient ID "K001" from the stamp information DB 144. The control unit 11 classifies the read stamp information into two types of stamp information indicating napping and other stamp information. The control unit 11 reads the patient's sleep record corresponding to the two types of stamp information indicating napping from the care content DB 145. The control unit 11 reads the care content of healthcare workers for the patient related to the other stamp information from the care content DB 145. The control unit 11 generates a prompt 60 by combining the patient's background, the care content of healthcare workers for the patient, the patient's sleep state during the first and second periods, the healthcare worker's records for the patient, and a command to generate advice regarding the patient.

[0097] Figure 22 is an explanatory diagram showing an example of a prompt. Prompt 60 includes a background field 41, a care content field 61, a sleep state field 42, a record field 43, and a generation command field 44. The contents described in the background field 41, the care content field 61, the sleep state field 42, and the generation command field 44 are the same as in Embodiment 3. The record field 43 contains the medical professional's record for the patient, as well as the patient's sleep record extracted from the record and the patient's sleep record identified from the stamp information.

[0098] In the record section 43 of Figure 22, the following has been added: "The previous day (April 10, 2024), the patient complained of insomnia and watched television in a chair in the dining room from around 8 PM. After watching television, the patient slept for about 3 hours in the chair. After that, the patient returned to their room and slept. On April 9, 2024, the patient dozed off in a wheelchair from 1:30 PM to 2:45 PM. On April 8, 2024, the patient dozed off in a chair, etc., from 12:08 PM to 2:05 PM."

[0099] The control unit 11 inputs the generated prompt 60 to the language model 12M, thereby obtaining the patient-related advice generated by the language model 12M. The subsequent processing is the same as in Embodiment 1, so the explanation is omitted.

[0100] According to Embodiment 3, the information processing system can receive a selection of target stamp information from multiple stamp information sources and obtain the care details of the medical professional corresponding to the received stamp information.

[0101] According to Embodiment 3, the information processing system can output advice regarding the patient by providing the acquired care details, sleep status during the first and second periods, the patient's background, and the medical professional's records regarding the patient to a language model.

[0102] (Embodiment 4) Embodiment 4 describes an example in which an information processing system outputs a second piece of advice about a patient by providing a language model with first advice about the patient, in addition to the patient's sleep state during the first and second periods, the patient's background, and records of healthcare professionals' interactions with the patient.

[0103] The first advice regarding the patient (hereinafter referred to as "first advice") is output by providing the language model with the patient's sleep status during the first and second periods, the patient's background, and the healthcare worker's records regarding the patient. The second advice regarding the patient (hereinafter referred to as "second advice") is output by providing the language model with the patient's sleep status during the first and second periods, the patient's background, the healthcare worker's records regarding the patient, and the first advice.

[0104] Figure 23 is a block diagram showing an example of a server configuration. The large-capacity storage unit 14 stores a reference information file 146. The reference information file 146 stores reference information data. The reference information data includes, for example, guidelines from domestic and international dementia societies, medical glossaries, medical papers, or professional journals on nursing care. The reference information data also includes data published on the internet, data converted from professional books into text files, data created independently by medical institutions, or data created independently by medical professionals. The reference information file 146 stores both the original reference information data and vectorized reference information data.

[0105] The method for creating the reference information file 146 is described below. The control unit 11 reads the reference information data via the reading unit 15. The reference information data may also be read by an external device (not shown). In that case, the control unit 11 receives the reference information data transmitted from the external device via the communication unit 13. The control unit 11 divides the read reference information data into chunks. A chunk is a component of the reference information data, such as a document, paragraph, sentence, phrase, and word. The control unit 11 vectorizes the divided reference information data. The control unit 11 stores the combination of the original reference information data and the vectorized reference information data in the reference information file 146. The control unit 11 creates the reference information file 146 by repeating the above process.

[0106] The processing of Embodiment 4 will now be described. In the following description, an example will be given in which the control unit 31 acquires the patient ID "K001". The control unit 31 sends the acquired patient ID "K001" and the startup notification to the server 10.

[0107] The control unit 11 receives the patient ID "K001" and activation notification transmitted from the smartphone 30. The control unit 11 reads the patient background corresponding to patient ID "K001" from the patient DB 141. The control unit 11 reads the records of healthcare workers for the patient associated with patient ID "K001" from the record DB 142. The control unit 11 reads the sensor data for the past week corresponding to patient ID "K001" from the sensor DB 143. Based on the sensor data for the past week that has been read, the control unit 11 identifies the sleep state of the patient during the first and second periods.

[0108] The control unit 11 generates a prompt 40 by combining the patient's sleep state during the first and second periods, the patient's background, the medical professional's records regarding the patient, and a command to generate advice about the patient. The control unit 11 inputs the generated prompt 40 to the language model 12M to obtain the first advice generated by the language model 12M. The control unit 11 extracts specific keywords from the patient's background, the medical professional's records regarding the patient, the command to generate advice about the patient, or the first advice. The specific keywords are, for example, medical keywords such as dementia, fracture, repositioning, or grooming. The specific keywords are pre-set and can be changed as appropriate according to the embodiment. The control unit 11 may extract specific sentences instead of specific keywords.

[0109] The control unit 11 vectorizes the extracted keywords. The control unit 11 refers to the reference information file 146 and calculates the similarity between the vectorized keywords and the vectorized reference information data. For example, cosine similarity or k-nearest neighbors is used to calculate the similarity between vectors. The control unit 11 identifies a portion of the vectorized reference information data whose calculated similarity is above a predetermined threshold. The control unit 11 reads a portion of the original reference information data (hereinafter referred to as "reference information") corresponding to the identified portion of reference information data from the reference information file 146. The reference information is a portion of the original reference information data and assists in the output of the second advice by the language model 12M.

[0110] The control unit 11 generates a prompt 70 by combining a first piece of advice, the patient's sleep state during the first and second periods, the patient's background, the medical professional's record of the patient, reference information, and a command to generate the second piece of advice. Figure 24 is an explanatory diagram showing an example of a prompt. The prompt 70 includes a background field 41, a sleep state field 42, a record field 43, a first piece of advice field 71, a reference information field 72, and a second piece of generation command field 73. The contents described in the background field 41, the sleep state field 42, and the record field 43 are the same as in Embodiment 1.

[0111] The first advice column 71 contains the first advice output by the language model 12M. The first advice column 71 in Figure 24 contains the advice regarding the patient shown in Figure 8. The reference information column 72 contains the reference information read from the reference information file 146. The reference information column 72 in Figure 24 contains the following: "A fracture refers to a bone breaking due to a strong external force. This is especially common in the elderly... Dementia is characterized by a decline in memory, judgment, and thinking ability..."

[0112] The second generation command field 73 contains instructions for the language model 12M to output the second advice. The second generation command field 73 in Figure 24 contains the following: "As a medical and nursing specialist, you should generate the second advice to be conveyed to the medical professionals responsible for A's care. When generating the second advice, please refer to the first advice. When generating the second advice, please consider the diseases and illnesses that A should be aware of based on the current situation, points that A should be aware of in daily life, and points that medical professionals should be aware of today."

[0113] The control unit 11 inputs the generated prompt 70 to the language model 12M, thereby obtaining the second advice generated by the language model 12M. The control unit 11 sends the obtained second advice to the smartphone 30. The control unit 31 receives the second advice sent from the server 10. The control unit 31 displays the received second advice on the display unit 35.

[0114] Figure 25 is a flowchart illustrating the program's processing flow. The flowchart in Figure 25 is a modified version of the process shown in Figure 10, with steps S206 onwards changed to steps S501-S509 and steps S601-S602. Steps similar to those in Figure 10 are omitted from explanation. The control unit 11 obtains the first advice by inputting prompts to the language model 12M, including the sleep state for the first and second periods, the patient's background, the medical professional's records regarding the patient, and a command to generate advice about the patient (step S501). The control unit 11 extracts specific keywords from the patient's background, the medical professional's records regarding the patient, the command to generate advice about the patient, or the first advice (step S502).

[0115] The control unit 11 vectorizes the extracted keywords (step S503). The control unit 11 refers to the reference information file 146 and calculates the similarity between the vectorized keywords and the vectorized reference information data (step S504). The control unit 11 identifies a portion of the vectorized reference information data whose calculated similarity is above a predetermined threshold (step S505). The control unit 11 reads the reference information corresponding to the identified portion of the reference information data from the reference information file 146 (step S506).

[0116] The control unit 11 generates a prompt 70 by combining the first advice, the patient's sleep state during the first and second periods, the patient's background, the medical professional's records regarding the patient, reference information, and a command to generate the second advice (step S507). The control unit 11 inputs the generated prompt 70 to the language model 12M and obtains the second advice generated by the language model 12M (step S508). The control unit 11 transmits the obtained second advice to the smartphone 30 (step S509).

[0117] The control unit 31 receives the second advice transmitted from the server 10 (step S601). The control unit 31 displays the received second advice on the display unit 35 (step S602).

[0118] According to Embodiment 4, the information processing system can output the second advice by generating a prompt that includes a first piece of advice, sleep status during the first and second periods, the patient's background, records of medical professionals regarding the patient, reference information, and a command to generate the second piece of advice, and by providing the generated prompt to the language model.

[0119] (Embodiment 5) Embodiment 5 describes an example in which an information processing system obtains feedback on a first piece of advice and includes the obtained feedback in the prompt.

[0120] The control unit 11 obtains first advice by inputting a prompt 40 to the language model 12M, which includes the patient's sleep state during the first and second periods, the patient's background, records of healthcare professionals regarding the patient, and a command to generate advice about the patient. The control unit 11 transmits the obtained first advice to the smartphone 30.

[0121] The control unit 31 receives the first advice transmitted from the server 10. The control unit 31 displays the received first advice on the display unit 35. The control unit 31 determines whether a predetermined time (for example, 1 hour) has elapsed since the first advice was displayed. If the control unit 31 determines that the predetermined time has not elapsed since the first advice was displayed, it continues to display the first advice. If the control unit 31 determines that the predetermined time has elapsed since the first advice was displayed, it displays a feedback reception screen (not shown) for the first advice on the display unit 35. The control unit 31 may also display the feedback reception screen for the first advice on the display unit 35 at a predetermined timing (for example, 30 minutes before the end of work). The medical professional inputs the feedback for the first advice on the reception screen via the input unit 34. The control unit 31 acquires the feedback for the first advice entered by the medical professional. The control unit 31 transmits the acquired feedback to the server 10.

[0122] The control unit 11 receives feedback transmitted from the smartphone 30. The control unit 11 extracts specific keywords from the patient's background, the healthcare worker's records regarding the patient, the instruction to generate advice about the patient, the first advice, or the feedback on the first advice.

[0123] The control unit 11 vectorizes the extracted keywords. The control unit 11 refers to the reference information file 146 and calculates the similarity between the vectorized keywords and the vectorized reference information data. The control unit 11 identifies a portion of the vectorized reference information data whose calculated similarity is above a predetermined threshold. The control unit 11 reads the reference information corresponding to the identified portion of the reference information data from the reference information file 146.

[0124] The control unit 11 generates a prompt by combining the first advice, feedback on the first advice, the patient's sleep state during the first and second periods, the patient's background, the medical professional's records regarding the patient, reference information, and a command to generate the second advice. The generated prompt includes the patient's background, the patient's sleep state during the first and second periods, the patient's background, the medical professional's records regarding the patient, the first advice, reference information, and the command to generate the second advice, as shown in Figure 21, as well as feedback on the first advice. Subsequent processing is the same as in Embodiment 4, so the explanation is omitted.

[0125] Figure 26 is a flowchart illustrating the program's processing flow. The flowchart in Figure 26 is the same as the process shown in Figure 25, but with steps S502-S507 changed to steps S701-S708 and steps S801-S806. Steps similar to those in Figure 25 are omitted from explanation. The control unit 11 transmits the acquired first advice to the smartphone 30 (step S701).

[0126] The control unit 31 receives the first advice transmitted from the server 10 (step S801). The control unit 31 displays the received first advice on the display unit 35 (step S802). The control unit 31 determines whether a predetermined time (for example, 1 hour) has elapsed since the first advice was displayed (step S803). If the control unit 31 determines that the predetermined time has not elapsed since the first advice was displayed (step S803: NO), it continues to display the first advice. If the control unit 31 determines that the predetermined time has elapsed since the first advice was displayed (step S803: YES), it displays a feedback reception screen for the first advice on the display unit 35 (step S804). The control unit 31 obtains the feedback for the first advice entered by the medical professional (step S805). The control unit 31 sends the obtained feedback to the server 10 (step S806).

[0127] The control unit 11 receives feedback transmitted from the smartphone 30 (step S702). The control unit 11 extracts specific keywords from the patient's background, the healthcare worker's record of the patient, the instruction to generate advice about the patient, the first advice, or the feedback on the first advice (step S703).

[0128] The control unit 11 vectorizes the extracted keywords (step S704). The control unit 11 refers to the reference information file 146 and calculates the similarity between the vectorized keywords and the vectorized reference information data (step S705). The control unit 11 identifies a portion of the vectorized reference information data whose calculated similarity is above a predetermined threshold (step S706). The control unit 11 reads the reference information corresponding to the identified portion of the reference information data from the reference information file 146 (step S707). The control unit 11 generates a prompt by combining the first advice, feedback to the first advice, the patient's sleep state during the first and second periods, the patient's background, the medical professional's records regarding the patient, reference information, and the command to generate the second advice (step S708). The control unit 11 proceeds to step S508.

[0129] According to Embodiment 5, the information processing system can obtain feedback on the first advice and include the obtained feedback in the prompt.

[0130] The matters described in each of the embodiments described above can be combined with one another. Furthermore, the independent claims and dependent claims described in the claims can be combined with one another in any combination, regardless of the form of reference. In addition, although the claims use a form in which claims referencing two or more other claims (multi-claim form), the claims are not limited to this. A form in which multi-claims referencing at least one multi-claim (multi-multi-claim) may also be used.

[0131] The embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The scope of the invention is indicated by the claims, not in the sense described above, and all modifications are intended to be in the sense and scope equivalent to the claims. [Explanation of Symbols]

[0132] 10. Information Processing Equipment (Server) 11 Control Unit 12 Storage section 12P control program 13 Communications Department 14 Mass storage 141 Patient DB 142 Record DB 142a Record Sheet R1 Patient Data Section 143 Sensor DB 143a Sensor Sheet 144 Stamp Information Database 144a Stamp Information Sheet 145 Caregiving Content Database 146 Reference Information File 15 Reading section 1a Portable storage medium 20 Sensor I / F device 21 Bed Sensor 30 Terminal devices (smartphones) 31 Control Unit 32 Storage section 32P control program 33 Communications Department 34 Input section 35 Display section 40 Prompts 41 Background column 42 Sleep Status Section 43 Record Section 44 Generated instruction field 50 Prompts 51 Excretion Status Section 60 prompts 61. Caregiving details section 70 Prompts 71. First Advice Section 72 Reference information column 73 2nd generated instruction column

Claims

1. Based on sensor data obtained from sensors provided for each patient, the patient's condition, background, and records of medical personnel's interactions with the patient are acquired. By providing the acquired state, background, and records to a language model, advice regarding the patient is output. A program that instructs a computer to perform a process.

2. The aforementioned sensor is a bed sensor, and based on sensor data obtained from the bed sensor over a predetermined period, it identifies the sleep state of the patient over a predetermined period. The program according to claim 1.

3. The sensor is a bed sensor, and based on sensor data obtained from the bed sensor for a predetermined period, the sleep state is identified, including the type of sleep and the number of occurrences of each type during the first period of the patient, and the type of sleep and the number of occurrences of each type during the second period, which includes the first period. By providing the language model with the sleep state, background, and records from the first and second periods, advice regarding the patient is output. The program according to claim 1.

4. Based on the sensor data, the excretion status of the patient identified over a predetermined period is obtained. By providing the acquired sleep state, excretion state, background, and records to the language model, advice regarding the patient is output. The program according to claim 2 or 3.

5. Background information includes the patient's gender, age, or symptoms. The program according to any one of claims 1 to 3.

6. The record includes the date and time, and the records of the medical personnel involved at that time. The program according to any one of claims 1 to 3.

7. Stamp information indicating the care provided by medical professionals to patients is stored in memory. The system accepts the selection of target stamp information from multiple stamp information entries. The care details of the medical professional corresponding to the stamp information received are obtained. By providing the acquired status, background, records, and care details to the language model, advice regarding the patient is output. The program according to claim 1.

8. Obtain the first piece of advice about the patient output by the language model. By providing the acquired state, background, records, and the first advice to the language model, a second piece of advice regarding the patient is output. The program according to claim 1.

9. A prompt is generated that includes the first advice, the state, the background, the record, reference information to assist in the output of the second advice, and a command to generate the second advice. The second advice is output by providing the generated prompt to the language model. The program according to claim 8.

10. Advice regarding the patient is output to the terminal device of the healthcare professional. The program according to claim 1.

11. In addition to providing advice regarding the aforementioned patient, the sleep state and records are output to the terminal device. The program according to claim 10.

12. Obtain feedback on the first piece of advice mentioned above. Include the acquired feedback in the prompt. The program according to claim 9.

13. An information processing device having a control unit, The control unit, Based on sensor data obtained from sensors provided for each patient, the patient's condition, background, and records of medical personnel's interactions with the patient are acquired. By providing the acquired state, background, and records to a language model, advice regarding the patient is output. Information processing device.

14. Based on sensor data obtained from sensors provided for each patient, the patient's condition, background, and records of medical personnel's interactions with the patient are acquired. By providing the acquired state, background, and records to a language model, advice regarding the patient is output. Information processing methods.

Images