Bed control system

Abstract

To provide a bed control system capable of more efficiently performing bed control with a low burden.SOLUTION: A bed control system is provided with a bed allocation part 131 consisting of one or more learned models for outputting allocation data, the learned models being generated by machine learning using first teacher data including bed data being data related to bed information, medical worker data being data related to information of a medical worker, injury and illness data being data related to injury and illness, and patient data being data related to patient information, and second teacher data including data related to one or more beds suitable to be allocated to a patient or an attendant on the basis of the first teacher data, and outputting allocation data related to one or more beds suitable to be allocated to the patient or the attendant on the basis of first input data including the bed data, the medical worker data, the injury and illness data and the patient data.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a bed control system for allocating beds in a hospital to patients or attendants. [Background technology]

[0002] Bed control involves keeping track of bed usage, patient admissions and discharges, transfers to other rooms, departments, wards, and transfers, as well as the patient's condition and whether or not there is a companion, and efficiently managing beds to ensure that a bed is available for use by the patient or companion depending on the patient's condition and urgency.This is an important task for hospitals, but because various factors must be taken into consideration, the work time and effort required tends to become excessive, and it places a heavy burden on the hospital.

[0003] In view of this situation, systems have been proposed in recent years to support bed control in order to perform bed control more efficiently.For example, Patent Document 1 discloses a technology that displays information necessary for bed control, such as the status of patients in each bed in a hospital, the operating status of all beds, and the usage status of each ward and the entire hospital, in an easy-to-understand format. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 2019-101675 Summary of the Invention [Problem to be solved by the invention]

[0005] However, the technology disclosed in Patent Document 1 is merely a technology for supporting and streamlining bed control in hospitals, and the final bed allocation (final solution) must be determined manually, so it does not solve the essential problem. In other words, the final solution cannot be obtained immediately.

[0006] The present invention has been made in view of the above-mentioned conventional circumstances. One of the objects of the present invention is to provide a bed control system that can perform bed control in a hospital more efficiently and with less burden. Means for solving the problem and effects of the invention

[0007] A bed control system according to a first aspect of the present invention is a bed control system that allocates beds in a hospital to patients or attendants, and can be configured to include a bed allocation unit consisting of one or more trained models generated by machine learning using first training data including bed data, which is data related to information about beds, medical worker data, which is data related to information about medical workers, injury / illness data, which is data related to injuries / illnesses, and patient data, which is data related to information about patients, and second training data, which is data related to one or more beds suitable for allocation to the patient or the attendant based on the first training data, and which outputs allocation data, which is data related to one or more beds suitable for allocation to the patient or the attendant, based on first input data including the bed data, the medical worker data, the injury / illness data, and the patient data.

[0008] According to the above configuration, for example, by using a trained model generated by machine learning past bed control records, allocation data conforming to past bed control (such as trends, practices, and orientations of hospital bed allocation) can be obtained. That is, when a bed is presented based on the allocation data, the user only needs to confirm whether that bed is acceptable. When multiple beds are presented, the user only needs to select one from among them. This allows bed control to be performed more efficiently and with less burden. This makes bed control, which previously required a certain level of expertise, relatively easy even for those without such expertise. Thus, the bed control system according to the present invention is characterized by its ability to directly obtain a final solution, which is significantly different from conventional technologies that simply support and streamline bed allocation (without obtaining a final solution).

[0009] Furthermore, unlike when bed control is automated mechanically using mathematical models (calculation formulas), it will be possible to use machine learning to reproduce bed control based on reasons that are difficult to verbalize or formulate into rules, such as "tacit knowledge" in hospitals.

[0010] In a bed control system according to a second aspect of the present invention, the bed allocation unit can be configured to have a first shared trained model generated by machine learning using data from the first teacher data and the second teacher data that can be shared among different hospitals, and a first dedicated trained model generated by machine learning using data from the first teacher data and the second teacher data that cannot be shared among different hospitals.

[0011] According to the above configuration, different hospitals can share the first common trained model, thereby reducing the costs associated with generating trained models. Furthermore, the first common trained model does not necessarily need to be generated by machine learning using data that can be shared among all hospitals using the bed control system; it is understood that this also includes cases where the first common trained model can be shared only among some hospitals. In this case, it is rational because data can be shared between similar hospitals, and a higher learning effect can be expected than when machine learning is performed using data from a single hospital. Furthermore, by applying the first dedicated trained model to each hospital, bed control that reflects the unique rules and habits of each hospital, which cannot be covered by the first common trained model, can be achieved.

[0012] In a bed control system according to a third aspect of the present invention, the bed allocation unit can be configured to have an emergency trained model which is a trained model that includes emergency data, which is data related to whether or not there is an emergency, in the first teacher data and the first input data, and a peacetime trained model which is a trained model that does not include emergency data in the first teacher data and the first input data.

[0013] According to the above configuration, different bed controls can be realized in peacetime and emergency situations. Specifically, a peacetime trained model to be used in peacetime and an emergency trained model to be used in emergency situations can be prepared separately in advance, and bed control can be performed based on the peacetime trained model in peacetime (peacetime mode), and bed control can be performed based on the emergency trained model in emergency situations (emergency mode). Here, the emergency trained model can be generated by machine learning based on triage (determining treatment priorities according to urgency and severity), for example.

[0014] A bed control system according to a fourth aspect of the present invention can be configured to include a shortage calculation unit that outputs data relating to the shortage of beds or medical staff for treating an injury or illness, based on the first input data.

[0015] According to the above configuration, data on the shortage of beds or medical personnel can be output. For example, based on the data, the number of beds or medical personnel shortages can be communicated to the user, allowing the user to easily grasp the shortage of beds or medical personnel. The shortage calculation unit can also collect data on the shortage of beds or medical personnel for treating injuries and illnesses at multiple hospitals via the bed control system 1, and generate and output organized data. That is, based on the situations of multiple hospitals, it is possible to make a judgment such as, "Hospital A is short of beds, but Hospital B and Hospital C have available beds, so it is possible to transfer patients to other hospitals," or to provide information such as, "Hospital D is short of doctors specializing in a, but Hospital E has a doctor specializing in a who has no surgeries scheduled for today, so it is possible to send that doctor to provide support." This allows, for example, multiple pieces of information to be collected instantly in the event of an emergency, allowing for prompt decisions such as admitting or transferring patients or requesting additional doctors.

[0016] A bed control system according to a fifth aspect of the present invention can be configured to include a second input unit through which a user inputs second input data including the patient data in the form of text or voice, an input guidance unit that guides the user to input information necessary for the first input data as the second input data, and a language recognition unit consisting of one or more trained models that are generated by machine learning using third training data including a thesaurus of terms (synonyms and related words) and that converts the second input data into a form that can be used as the first input data and outputs it.

[0017] According to the above configuration, the user can input information required for the first input data in the form of text or voice by following the guidance of the input guidance section, and therefore the bed control system can be used easily.

[0018] In a bed control system according to a sixth aspect of the present invention, the language recognition unit can be configured to have a second shared trained model generated by machine learning using data from the third training data that can be shared among different hospitals, and a second dedicated trained model generated by machine learning using data from the third training data that cannot be shared among different hospitals.

[0019] According to the above configuration, the second common trained model can be shared among different hospitals, thereby reducing the costs associated with generating trained models. Furthermore, the second common trained model does not necessarily need to be generated by machine learning using data that can be shared among all hospitals using the bed control system; it is understood that this also includes cases where the model can be shared only among some hospitals. In this case, it is rational because data can be shared among similar hospitals, and a higher learning effect can be expected than when machine learning is performed using data from a single hospital. Furthermore, by applying the second dedicated trained model to each hospital, a bed control system can be realized that can handle hospital-specific terminology (such as jargon and abbreviations) that cannot be handled by the second common trained model.

[0020] A bed control system according to a seventh aspect of the present invention comprises: a display unit that, when the bed allocation unit outputs allocation data, which is data relating to multiple beds suitable for allocation to the patient or the attendant, displays the multiple beds in a list; a selection unit that allows a user to select one bed from the multiple beds displayed on the display unit; and a correction unit consisting of one or more trained models that is generated by machine learning using fourth teacher data including the data relating to the multiple beds and data relating to the bed selected by the selection unit, and outputs correction data, which is data for correcting the allocation data based on the habits of the user, and the display unit can be configured to display the list of the multiple beds that reflects the habits of the user based on the correction data. [Brief explanation of the drawings]

[0021] [Figure 1]1 is a system configuration diagram showing the configuration of a bed control system according to an embodiment of the present invention. [Figure 2] FIG. 2 is a functional block diagram showing functions of a bed control system according to an embodiment of the present invention. [Figure 3] 3 is a schematic diagram illustrating the processing of an input guidance unit and a correction unit of the bed control system according to the embodiment of the present invention. FIG. DETAILED DESCRIPTION OF THE INVENTION

[0022] Embodiments of the present invention will be described below with reference to the drawings. However, the embodiments described below are merely examples for embodying the technical concept of the present invention, and the present invention is not limited to these. Furthermore, this specification in no way specifies the components set forth in the claims as components of the embodiments. The dimensions, materials, shapes, and relative positions of components described in the embodiments are not intended to limit the scope of the present invention, and are merely illustrative unless otherwise specified. The size and relative positions of components shown in the drawings may be exaggerated for clarity. Furthermore, in the following description, the same names and symbols indicate identical or similar components, and detailed descriptions will be omitted as appropriate. Furthermore, each element constituting the present invention may be configured with the same component, so that multiple elements are served by a single component. Conversely, the function of a single component may be shared among multiple components. Furthermore, the functions or processing of each element constituting the present invention may be realized, to the extent possible, by artificial intelligence (AI). [Bed Control System 1]

[0023] A bed control system 1 according to one embodiment of the present invention will be described with reference to FIGS.

[0024] The bed control system 1 is a system for controlling beds in a hospital, and is made up of a network 11, a user terminal 12, and a server 13, as shown in FIG.

[0025] Note that the mode of the bed control system 1 is not limited to the above. For example, if the user terminal 12 is configured to have the functions of the server 13 described later, the bed control system 1 can be realized by the user terminal 12 alone. [Network 11]

[0026] As shown in FIG. 1, the network 11 is a communication system (a normal internet network) for connecting a plurality of user terminals 12 and a server 13 so that they can communicate with each other, and the form of the network is not particularly limited. [User terminal 12]

[0027] 1, the user terminal 12 is an information terminal owned by a hospital and used by a person who performs bed control work at the hospital (hereinafter referred to as a "user US"), and corresponds to various devices such as a personal computer, tablet, smartphone, etc. The user terminal 12 does not mean a specific information terminal, and when the user US uses multiple information terminals, all of them correspond to the user terminal 12.

[0028] In addition, the user terminal 12 corresponds to an example of the "second input unit," "display unit," and "selection unit" in the claims, and performs data input by operating a keyboard or touch panel, voice input into a microphone, etc., sending and receiving data with the server 13 via the network 11, and transmitting data to the user US. [Server 13]

[0029] As shown in FIG. 1, the server 13 is a server managed by an administrator AD, and is connected to a user terminal 12 via a network 11 so as to be able to communicate with the user terminal 12 .

[0030] As shown in FIG. 2, the server 13 includes a bed allocation unit 131, a shortage calculation unit 132, an input guidance unit 133, a language recognition unit 134, a correction unit 135, and a learning unit 136, and performs processing related to bed control based on data transmitted via the user terminal 12. Bed Allocation Department 131

[0031] The bed allocation unit 131 is a component consisting of a plurality of trained models generated by the learning unit 136 through machine learning using first training data including bed data, which is data related to information about beds, medical staff data, which is data related to information about medical staff, injury and illness data, which is data related to injuries and illnesses, and patient data, which is data related to information about patients, and second training data including data related to beds suitable for allocation to patients or attendants based on the first training data.

[0032] In addition, when actually performing bed control, the bed allocation unit 131 outputs allocation data, which is data related to one or more beds suitable for allocation to a patient or attendant, based on first input data including bed data, medical staff data, injury / illness data, and patient data of the patient to whom the bed is actually allocated at that time. First teacher data

[0033] The first training data is data used to generate a trained model that constitutes the bed allocation unit 131, and includes bed data, medical staff data, injury / illness data, patient data, and emergency data. In other words, when it is determined that a specific bed should be assigned to a specific patient or attendant, it is data related to information that is the reason or cause of the determination. The first training data may be data used in past bed control (such as actual patient data) or may be data (fictitious data) created for machine learning by the bed allocation unit 131.

[0034] Examples of data that may be included in the first teacher data are shown below.

[0035] Bed data: In addition to bed identification information, this may include data such as the bed's configuration, distance from the nurse's station, whether it is a shared room or a private room, and the type of room the bed is in. It may also include data such as the period when the bed is unavailable due to construction work at the hospital.

[0036] Healthcare worker data: This may include the identity of the healthcare worker, as well as data about what the healthcare worker can do (such as whether they can perform a particular procedure) and their attendance status.

[0037] Injury / Illness Data: In addition to identifying information about the injury or illness, this may include data about the details of the injury or illness, treatment methods (including medication, surgery, and other treatment data), the duration and cost of the treatment, etc.

[0038] Patient data: In addition to the patient's identification information, this may include data such as gender, age, financial status, details of the injury or illness, whether the patient has any other chronic illnesses, whether the patient is accompanied by an attendant, and the number of attendants.

[0039] Emergency data: This may include information on whether the situation is an emergency or peacetime, and the details of the emergency (for example, earthquakes, pandemics, wartime, etc.). Emergency data is prepared in advance based on the knowledge of experts. Second training data

[0040] The second training data is data used to generate the trained model that constitutes the bed allocation unit 131, and includes identification information for one or more beds that are suitable for allocation to a patient or attendant when based on the first training data.

[0041] Specifically, if the first teacher data is data used in past bed control, the identification information of the actually assigned bed can be used. Also, if the first teacher data is data created for machine learning by the bed allocation unit 131, one or more beds suitable for allocation to the patient or attendant are selected based on the first teacher data, and the identification information of the bed is created.

[0042] In other words, the second teacher data can be said to be a judgment (answer for bed control) of a person performing bed control based on the first teacher data. The following cases (examples) are considered as examples of such judgment.

[0043] Case 1: The patient is a disabled infant, so a bed is assigned that can accommodate an attendant bed next to it (the attendant is assigned an attendant bed next to it).

[0044] Case 2: The patient is female, so she is assigned a bed with no male occupants.

[0045] Case 3: The patient makes strange noises, so other patients are not assigned to other beds in the same room.

[0046] Case 4: During the summer vacation period, the number of students admitted to hospital increases, so the length of hospital stay is appropriately shortened and beds are allocated to increase the turnover rate.

[0047] Case 5: The patient is financially well-off and desires higher quality service, so a bed in a private room (special room) is assigned.

[0048] Case 6: During the recovery period immediately after surgery, the patient's condition needs to be closely monitored, so a bed near the nurse's station is assigned.

[0049] Case 7: Because it is an emergency, beds are allocated based on triage (determining treatment priorities based on the urgency and severity of injuries and illnesses).

[0050] Case 8: Allocate beds to minimize bed and room changes during the hospital stay.

[0051] As mentioned above, cases 1 to 8 are possible judgments made by bed control personnel, but in reality, there are usually multiple conditions to consider that are intricately intertwined, and the judgment of the bed control personnel and the reasons and causes for that judgment do not necessarily correspond one-to-one.

[0052] In such cases, it is believed that the person in charge of bed control unconsciously considers a variety of complex conditions simultaneously, and then decides (judges) which conditions to prioritize (weight) before making a bed allocation decision.

[0053] In the bed control system 1 according to one embodiment of the present invention, such judgments are made by a bed allocation unit 131 generated by machine learning using first teacher data and second teacher data, thereby enabling the mechanization and reproduction of judgments that are made unconsciously (judgments that are difficult to verbalize or formulate into rules). First shared trained model 131a and first dedicated trained model 131b

[0054] In addition, the multiple trained models provided by the bed allocation unit 131 can be classified into a first shared trained model 131a generated by machine learning using data that can be shared among different hospitals, out of the first teacher data and second teacher data, and a first dedicated trained model 131b generated by machine learning using data that cannot be shared among different hospitals.

[0055] For example, the first shared trained model 131a can be generated using injury data, emergency data, etc. that can be shared among different hospitals, and the first dedicated trained model 131b can be generated using bed data, medical worker data, patient data, etc. that cannot be shared among different hospitals.

[0056] Furthermore, the first shared trained model 131a does not necessarily have to be generated by machine learning using data that can be shared among all hospitals that use the bed control system 1. For example, it is understood that this also includes cases where the data can be shared only among some hospitals, such as between hospitals of similar size (number of beds), between hospitals with similar hospital characteristics (strong in acute care, cancer specialist, cardiovascular specialist, organ transplant specialist, etc.), or between hospitals with similar regional characteristics (such as an area where lifestyle-related diseases are common because sugar and salt intake is higher than the national average). In this case, it is rational because data can be shared between similar hospitals, and a higher learning effect can be expected than by performing machine learning using data from a single hospital.

[0057] In this way, by providing the first common trained model 131a and the first dedicated trained model 131b independently, the first common trained model 131a can be shared among different hospitals, thereby reducing the cost of generating trained models. Furthermore, by applying the first dedicated trained model 131b to each hospital, bed control that reflects the unique rules and habits of each hospital can be realized. <Peacetime trained model 131H and emergency trained model 131U>

[0058] Furthermore, the multiple trained models provided in the bed allocation unit 131 can be classified into ordinary time trained models 131H that are used in ordinary times and emergency time trained models 131U that are used in emergency times.

[0059] The emergency trained model 131U differs from the peacetime trained model 131H in that the first training data used to generate it includes emergency data, and the second training data includes a judgment of Case 7.

[0060] In this way, by preparing separate models in advance for peacetime training to be used in peacetime and emergency training to be used in emergency situations, bed control can be performed based on the peacetime training model in peacetime (peacetime mode), and bed control can be performed based on the emergency training model in emergency situations (emergency mode). <<First input data>>

[0061] The first input data is data that is input into the trained model that constitutes the bed allocation unit 131 when actually performing bed control, and the contents of the data may include bed data, medical staff data, injury / illness data, patient data, and emergency data, similar to the first training data.

[0062] The method of inputting the input data is not particularly limited, but for example, among bed data, medical staff data, injury / illness data, and patient data, data that is known in advance, such as gender, age, and financial status, can be input in advance or obtained in cooperation with an external system such as an electronic medical record, and data among patient data that is determined just before bed allocation, such as bed contents and emergency data, can be input at that time. Allocation Data

[0063] The allocation data is data output from the trained model that constitutes the bed allocation unit 131, and is data related to one or more beds that are suitable for allocation to a patient or attendant, output based on the first input data, and may include data such as the identification information of the bed and the allocation period.

[0064] The output allocation data is transmitted to the user terminal 12 and communicated to the user US. The method of communication to the user US is not particularly limited, but may be displayed on a display in the form of characters or figures, or communicated by voice. For example, ct2-6 to ct2-8 in Fig. 3 are a list of identification information of a plurality of beds suitable for allocation to patients or attendants in the form of characters. Other Aspects of the Bed Allocation Unit 131

[0065] The mode of the bed allocation unit 131 is not limited to the above.

[0066] For example, the output of allocation data does not need to be solely the responsibility of the trained model constituting the bed allocation unit 131, and part of the responsibility may be shared with other methods such as a mathematical model. For example, the trained model is responsible for parts that are difficult to process mechanically (decisions that are difficult to verbalize or formulate as rules), and the mathematical model is responsible for parts that can be processed mechanically, thereby making it possible to output allocation data more efficiently.

[0067] Furthermore, it is not necessary for all trained models for outputting allocation data to be provided on the server 13, and some or all of them may be provided on the user terminal 12. For example, the first shared trained model 131a may be provided on the server 13, and the first dedicated trained model 131b may be provided on the user terminal 12. In other words, decisions that can be shared among different hospitals can be made on the server 13, and parts that require unique decisions for each hospital can be made on the user terminal 12 for each hospital, thereby reducing the costs associated with operating the system.

[0068] Furthermore, the trained models constituting the bed allocation unit 131 may be prepared by separating trained models for each season in accordance with the prevalence of illnesses and injuries in each season.

[0069] Furthermore, the trained models constituting the bed allocation unit 131 do not necessarily need to be multiple, and one trained model may be used to output allocation data (to make all decisions). <Shortage calculation unit 132>

[0070] The shortage calculation unit 132 is a component that outputs data on the shortage of beds or medical personnel for treating an injury or illness based on the first input data. This allows the number of shortage of beds or medical personnel to be communicated to the user US based on the data, for example, by displaying it on the user terminal 12, and the user US can easily understand the shortage of beds or medical personnel based on the display.

[0071] Furthermore, the shortage calculation unit 132 can collect data related to the shortage of beds or medical personnel for treating injuries and illnesses at multiple hospitals via the bed control system 1, and generate and output organized data (corresponding to an example of "data related to the shortage of beds or medical personnel for treating injuries and illnesses" in the claims). That is, based on the situations of multiple hospitals, it is possible to make decisions such as "Hospital A has a shortage of beds, but Hospital B and Hospital C have available beds, so it would be better to transfer the patient to another hospital," or "Hospital D has a shortage of doctors specializing in a, but Hospital E has a doctor specializing in a who has no surgeries scheduled for today, so it would be better to send that doctor to provide backup." This allows, for example, in the event of an emergency, multiple pieces of information to be collected instantly, and it is possible to quickly transfer patients to another hospital or request backup doctors.

[0072] Note that the mode of the deficiency calculation unit 132 is not limited to the above. For example, the method of outputting data from the deficiency calculation unit 132 is not particularly limited, and may use a mathematical model, a trained model, or a combination of these. <Input guidance section 133>

[0073] The input guidance unit 133 is a component that guides the user US to input information required for the first input data as second input data when the user US inputs data using the user terminal 12. The second input data is data whose content is the same as the first input data but whose format is different from that of the first input data. More specifically, the second input data may include bed data, medical staff data, injury / illness data, patient data, and emergency data, and its format may be text or audio.

[0074] When the user US uses the user terminal 12 to input second input data in the form of a sentence, the input guiding unit 133 communicates with the user US in a dialogue format (chat format) using natural language, for example, as shown in Fig. 3, outputs sentences that elicit information necessary for the first input data, and displays them on the user terminal 12. In Fig. 3, ct1-1 to ct1-6 are displays corresponding to the second input data input by the user US, and ct2-1 to ct2-9 display the sentences output by the input guiding unit 133.

[0075] For example, in Figure 3, ct2-1 is a sentence to find out the patient's gender, ct2-3 is a sentence to find out the reason why an escort is needed, and ct2-4 is a sentence to find out the gender of the escort.

[0076] The input guidance unit 133 may be implemented in any manner other than as described above. For example, by using voice synthesis technology, the chat-style dialogue described above may be replaced with a voice dialogue. The same applies when the user US inputs second input data in the form of voice using the user terminal 12. The means for determining the information required for the first input data and for realizing a dialogue in natural language are not particularly limited, and may use a mathematical model, a trained model, or a combination of these. <Language recognition unit 134>

[0077] This component is composed of multiple trained models that are generated by machine learning using third training data including a thesaurus of terms, and converts second input data entered by a user US in the form of text or voice using a user terminal 12 into a form that can be used as first input data and outputs it.

[0078] For example, as explained in the section on the input guidance unit 133, when second input data is received from the user US in a natural language interactive format, the second input data contains both information necessary as first input data and information not necessary. In such a case, the language recognition unit 134 extracts only the information necessary as first input data from the second input data and performs processing to match the extracted data with the format of the first input data. "Third training data"

[0079] The third training data is data used to generate the trained model that constitutes the language recognition unit 134, and includes a thesaurus of terms (terms classified and organized according to superordinate / subordinate relationships, part / whole relationships, synonymous relationships, similar relationships, etc.).

[0080] Examples of data that may be included in the third training data are shown below.

[0081] Thesaurus of general terms: A thesaurus of general terms as used in everyday conversation may be included.

[0082] Thesaurus of technical terms: A thesaurus of technical terms such as medical terms, nursing terms, etc. may be included.

[0083] Thesaurus of hospital-specific terminology: May include a thesaurus of hospital-specific terminology (slang, abbreviations, etc.). <<Second shared trained model 134a and second dedicated trained model 134b>>

[0084] In addition, the multiple trained models provided by the language recognition unit 134 can be classified into a second shared trained model 134a generated by machine learning using data from the third training data that can be shared among different hospitals, and a second dedicated trained model 134b generated by machine learning using data that cannot be shared among different hospitals.

[0085] For example, the second shared trained model 134a can be generated using a thesaurus of general terms or a thesaurus of specialized terms, and the second dedicated trained model 134b can be generated using a thesaurus of terms specific to the hospital.

[0086] Furthermore, the second shared trained model 134a does not necessarily have to be generated by machine learning using data that can be shared among all hospitals that use the bed control system 1. For example, this also includes cases where the data can be shared only among some hospitals, such as between hospitals of similar size (number of beds), between hospitals with similar hospital characteristics (strong in acute care, cancer specialist, cardiovascular specialist, organ transplant specialist, etc.), or between hospitals with similar regional characteristics (such as an area where sugar and salt intake is higher than the national average and lifestyle-related diseases are common). In this case, it is rational because data can be shared between similar hospitals, and a higher learning effect can be expected than by performing machine learning using data from a single hospital.

[0087] In this way, by providing the second common trained model 134a and the second dedicated trained model 134b independently, the second common trained model 134a can be shared among different hospitals, thereby reducing the cost of generating trained models. Also, by applying the second dedicated trained model 134b to each hospital, a bed control system that can handle terminology unique to each hospital (such as slang and abbreviations) can be realized. Other Aspects of the Language Recognition Unit 134

[0088] The language recognition unit 134 is not limited to the above embodiment.

[0089] For example, the conversion of the second input data into a form that can be used as first input data does not need to be performed solely by the trained model that constitutes the language recognition unit 134, but part of this task may be shared by other methods such as mathematical models.

[0090] Furthermore, it is not necessary to provide all of the trained models for the conversion in the server 13, and some or all of them may be provided in the user terminal 12. For example, the second shared trained model 134a may be provided in the server 13, and the second dedicated trained model 134b may be provided in the user terminal 12. In other words, decisions that can be shared among different hospitals can be made in the server 13, and decisions that require unique decisions for each hospital can be made in the user terminal 12 for each hospital, thereby reducing the costs associated with operating the system.

[0091] Furthermore, the trained models constituting the bed allocation unit 131 do not necessarily need to be multiple, and one trained model may be used to output allocation data (to make all decisions).

[0092] Furthermore, if the first input data (or data in a similar format) can be obtained from the beginning by using a dedicated input form (which corresponds to an example of the "input guidance section" in the claims), the language recognition section 134 does not necessarily need to be provided. <Correction unit 135>

[0093] The correction unit 135 is a component that learns the habits of the user US when selecting one bed from multiple beds suitable for allocation to a patient or attendant, and performs corrections based on those habits during subsequent bed control.

[0094] For example, if there are three beds suitable for allocation to a patient or attendant, the bed control system 1 lists the identification information of the three beds along with their numbers, as shown in ct2-6 to ct2-8 in Fig. 3. At this time, the higher the degree of suitability for allocation to a patient or attendant (hereinafter referred to as "suitability"), the smaller the number. In other words, the bed allocation unit 131 determines that the bed marked with "1." is the most suitable bed.

[0095] In response to this, the user US selects the bed that he or she actually wants to assign to the patient or attendant, as shown in ct1-6 in Figure 3, and the correction unit 135 learns the fourth teacher data including data related to this selected bed, the first input data, and the assignment data.

[0096] In other words, even if there are multiple beds suitable for allocation by a patient or attendant, if the bed allocation unit 131 could make exactly the same judgment as the user US, the user US would always select the bed marked with "1." by the bed allocation unit 131, but in reality this is not the case. The correction unit 135 learns such differences between the judgments of the user US and the judgments of the bed allocation unit 131, and outputs correction data that is data for correcting this difference.

[0097] Then, in subsequent bed control, by using this correction data, it is possible to display a plurality of beds in a manner that reflects the habits of the user US.

[0098] The form of the correction unit 135 is not limited to the above. Also, the bed control system 1 does not necessarily have to include the correction unit 135. <Study Section 136>

[0099] The learning unit 136 is a component included in the server 13 that generates a trained model.

[0100] There are no particular limitations on the mode of the learning unit 136 or the machine learning algorithm. By periodically regenerating (re-learning) the trained model, the accuracy of the trained model can be improved.

[0101] Furthermore, the learning unit 136 does not necessarily have to be provided only in the server 13, but may be provided in the user terminal 12, or may not be provided within the bed control system 1 (in which case the trained model is generated outside the system). [Effects of the present invention]

[0102] As described above, the bed control system 1 according to one embodiment of the present invention can acquire allocation data conforming to past bed control, for example, by using a trained model generated by machine learning past bed control records. That is, when a bed is presented based on the allocation data, the user US only needs to confirm whether that bed is acceptable, and when multiple beds are presented, the user US only needs to select one bed from among them, thereby enabling bed control to be performed more efficiently and with less burden.

[0103] Furthermore, unlike automating bed control using mathematical models, etc., it will be possible to learn and reproduce bed control based on reasons that are difficult to verbalize or formulate into rules. [Explanation of symbols]

[0104] 1. Bed control system 11. Network 12...User terminal 13...Server 131...bed allocation unit; 131a...first shared trained model; 131b...first dedicated trained model; 131H...normal time trained model; 131U...emergency time trained model 132...Shortage calculation section 133...Input guidance section 134... language recognition unit; 134a... second shared trained model; 134b... second dedicated trained model 135...correction unit 136…Study Department US...User AD…Administrator

Claims

1. A bed control system for allocating beds in a hospital to patients or attendants, comprising: a bed allocation unit including one or more trained models generated by machine learning using first training data including bed data, which is data related to information on beds, medical staff data, which is data related to information on medical staff, injury and illness data, which is data related to injury and illness, and patient data, which is data related to information on patients, and second training data, which is data related to one or more beds suitable for allocation to the patient or the attendant based on the first training data, and which outputs allocation data, which is data related to one or more beds suitable for allocation to the patient or the attendant, based on first input data including the bed data, the medical staff data, the injury and illness data, and the patient data; Equipped with bed control system.

2. 2. A bed control system according to claim 1, The bed allocation unit a first shared trained model generated by machine learning using data that can be shared among different hospitals, out of the first teacher data and the second teacher data; a first dedicated trained model generated by machine learning using data that cannot be shared among different hospitals, out of the first teacher data and the second teacher data; A bed control system having:

3. 2. A bed control system according to claim 1, The bed allocation unit An emergency trained model that is a trained model including emergency data that is data related to whether or not the first teacher data and the first input data are an emergency; a peacetime trained model that is a trained model that does not include emergency data in the first teacher data and the first input data; A bed control system having:

4. 10. The bed control system of claim 1, further comprising: A bed control system including a shortage calculation unit that outputs data related to the shortage of beds or medical personnel for treating an injury or illness based on the first input data.

5. 10. The bed control system of claim 1, further comprising: a second input unit through which a user inputs second input data including the patient data in the form of text or voice; an input guidance unit that guides the user to input information necessary for the first input data as the second input data; a language recognition unit including one or more trained models generated by machine learning using third teacher data including a thesaurus of terms, which converts the second input data into a form usable as the first input data and outputs the converted data; A bed control system comprising:

6. 6. A bed control system according to claim 5, The language recognition unit A second shared trained model generated by machine learning using data that can be shared among different hospitals among the third teacher data; and A second dedicated trained model generated by machine learning using data that cannot be shared among different hospitals among the third teacher data; and A bed control system having:

7. 10. The bed control system of claim 1, further comprising: a display unit that displays a list of a plurality of beds suitable for allocation to the patient or the attendant when the bed allocation unit outputs allocation data that is data related to the plurality of beds; a selection unit that allows a user to select one bed from among the plurality of beds displayed on the display unit; a correction unit including one or more trained models generated by machine learning using fourth teacher data including data related to the plurality of beds and data related to the bed selected by the selection unit, and outputting correction data that is data for correcting the assignment data based on the habits of the user; and Equipped with A bed control system in which the display unit displays a list of the plurality of beds that reflects the user's habits based on the correction data.

Images