System and method for predicting call times for specimen collection

Abstract

Provided are sample collection call time prediction system and method, with which it is possible to improve the accuracy of predicting the time at which a patient is called for sample collection. This call time prediction system 10 comprises a first processor (prediction unit 101). The first processor predicts, by machine learning, the time at which a patient is called for sample collection, the prediction being made on the basis of at least one of reception time information (reception time 302) indicating the reception time for a patient from whom a sample is to be collected, sample type information 303 indicating the type of the sample to be collected from the patient, a reception number 304 indicating the order of reception of the patient, inpatient / outpatient classification information 305 indicating whether the patient is an inpatient or an outpatient, and the number of patients (number of waiting patients 306) waiting to be called for sample collection at the reception time 302.

Description

Technical Field

[0001] The present invention relates to a system and method for predicting the calling time of specimen collection.

Background Art

[0002] A specimen collection support system is related to the blood collection work in particular in a medical institution, and supports the specimen collection work by automating and centrally managing peripheral operations of specimen collection such as patient reception for specimen collection, patient calling for specimen collection, preparation of specimen containers according to examination requests, barcode attachment to specimen containers, and end management of specimen collection.

[0003] Normally, a number is assigned to a patient at the same time as the reception of specimen collection, and the patient understands that their turn has come by receiving a call according to that number.

[0004] A general specimen collection support system does not provide information to the patient such as when the call will be made at the time of reception. Therefore, unless the medical institution staff provides information to the patient about approximately when they will be called based on their ingenuity, or the patient predicts based on their own experience when they are likely to be called, the patient will wait in the waiting room without knowing when they will be called.

[0005] If it is a non-congested time zone, the call will be made in about a few minutes, but if it is a congested time zone in the morning, the patient may wait in the waiting room for more than 30 minutes, which is a great burden. In contrast, a system for predicting and calculating the blood collection waiting time is known (for example, see Patent Document 1).

Prior Art Documents

Patent Documents

[0006]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0007] In technologies such as those disclosed in Patent Document 1, the accuracy of predicting waiting times decreases due to changes in the operation of sample collection procedures, etc.

[0008] The present invention aims to provide a call time prediction system and method that can improve the accuracy of predicting the time when a patient will be called in for sample collection. [Means for solving the problem]

[0009] To achieve the above objective, an example of the call time prediction system of the present invention includes a storage unit that stores reception time information indicating the reception time of multiple patients from whom specimens are to be collected, specimen type information indicating the type of specimen to be collected from the multiple patients, reception number indicating the reception order of the multiple patients, inpatient / outpatient classification information indicating whether the multiple patients are inpatients or outpatients, the number of patients waiting to be called for specimen collection at the reception time, and the measured values ​​of multiple call times for which the multiple patients were called, and the reception time information for each of the multiple patients stored in the storage unit, and the call time for each of the multiple patients, Specimen type information indicating the type of specimen to be collected from the aforementioned multiple patients, and the number of patients waiting to be called for specimen collection during the reception time, We build a machine learning model using the following as input data, at least Specimen type information indicating the type of specimen to be collected from the aforementioned multiple patients, and the number of patients waiting to be called for specimen collection during the reception time, A sample collection call time prediction system includes a prediction unit that predicts the time at which the multiple patients will be called for sample collection based on the machine learning model, and an evaluation unit that evaluates the time at which the multiple patients will be called for sample collection as predicted by the prediction unit. system And, The evaluation unit, The machine learning model is rebuilt at a specified time during the day, and the learning period is set to multiple learning periods that vary at regular intervals within the specified period. A provisional machine learning model is constructed for each learning period. The difference between the predicted time when the patient is called, as predicted by the evaluation unit, and the actual measured call times is calculated for each of the multiple learning periods. The learning period of the provisional machine learning model that has the highest probability of the calculated difference being within a threshold, which is an indicator of prediction accuracy, is determined. After setting the determined learning period, the machine learning model used in the business is rebuilt. [Effects of the Invention]

[0010] According to the present invention, the accuracy of predicting the time when a patient will be called for sample collection can be improved. Problems, configurations, and effects other than those described above will be clarified by the following description of embodiments. [Brief explanation of the drawing]

[0011] [Figure 1] This is a schematic diagram of a sample collection call time prediction system that takes into account the periodic updating of machine learning models. [Figure 2] This information is used to predict the call time for sample collection using a machine learning model. [Figure 3] This information, used in conjunction with Figure 2, is necessary for building machine learning models. [Figure 4] This is the parameter setting screen for building a machine learning model. [Figure 5] This is a specimen collection registration form issued to patients at the time of registration for specimen collection. [Figure 6] This is a monitor screen that displays a list of patients waiting to be called after their sample collection has been registered. [Figure 7] This is a schematic diagram of a sample collection call time prediction system, keeping in mind that the machine learning model will not be updated regularly. [Figure 8] This is a schematic diagram of a system for predicting call times for specimen collection, which includes a function for managing entry into the specimen collection waiting room. [Modes for carrying out the invention]

[0012] Hereinafter, with reference to the drawings, the configuration and operation of the call time prediction system according to the first to third embodiments of the present invention will be described. One of the problems to be solved in this embodiment is to enable the effective utilization of the waiting time of the patient from the acceptance of the specimen collection to the call. In this embodiment, by presenting the predicted time of when the call will be made at the time of accepting the specimen collection, it encourages the patient to effectively utilize the waiting time. Another problem to be solved in this embodiment is to avoid congestion in the specimen collection waiting room. By presenting the predicted call time for specimen collection in advance, it guides patients to gather at the specimen collection location when the call time approaches. This contributes to avoiding the three Cs required for preventing the spread of the novel coronavirus infection these days, and aims to reduce the infection risks for medical staff and patients.

[0013] (First Embodiment) Hereinafter, an embodiment of the present invention will be described with reference to FIG. 1.

[0014] FIG. 1 is an overall schematic diagram of a specimen collection call time prediction system 10 according to the first embodiment of the present invention, with the idea of periodically updating the machine learning model in mind.

[0015] The call time prediction system 10 includes a reception unit 102 responsible for receiving patients for specimen collection, a call unit 104 for notifying the patient that it is their turn for specimen collection, a storage unit 103 for storing information related to patient reception and calls, a prediction unit 101 for predicting the time from specimen collection reception to the call by machine learning, and a display unit 105 for monitoring and displaying (for example, displaying on a browser) the prediction result.

[0016] The specimen collection call time prediction system 10 of the present invention is assumed to present the predicted call time for specimen collection to the patient by cooperating with or being a part of the specimen collection support system 20.

[0017] When cooperating with the specimen collection support system 20, the reception unit 102 and the calling unit 104 belong to the side of the specimen collection support system 20, and the storage unit 103 and the prediction unit 101 belong to the side of the call time prediction system 10. The specimen collection support system 20 and the call time prediction system 10 are connected via a medical facility internal LAN (Local Area Network) or the Internet, and exchange information with each other. The display unit 105 belongs to the side of the call time prediction system 10, but the installation location is assumed to be near the reception unit 102 or the calling unit 104, and data is obtained from the storage unit 103 via a medical facility internal LAN or the Internet.

[0018] The patient who has received the instruction for specimen collection moves to the specimen collection location and operates the installed reception unit 102 to register himself / herself in the waiting queue for specimen collection. Depending on the medical facility, the retrieval of the medical record may be linked to the reception of specimen collection. In this case, the reception unit 102 automatically registers the patient in the waiting queue for specimen collection upon receiving an instruction from the medical record system.

[0019] In conjunction with the registration in the waiting queue, the reception unit 102 transmits the reception information 110 to the storage unit 103 and notifies that the patient has been received.

[0020] FIG. 2 shows a list of data related to the reception unit 102 obtained by the storage unit 103 through the reception information 110. The reception information 110 includes at least the patient identification information 301 for identifying the patient who has received the specimen collection.

[0021] The patient identification information 301 preferably uses a dual key of the patient ID or the reception number and the date. It is necessary to be able to link with the data of the same patient received from the calling unit 104 later and to distinguish the data of the same patient on other days.

[0022] One piece of information that the memory unit 103 must obtain is the sample collection reception time 302. The sample collection reception time 302 includes the year, month, day, hour, minute, and second information of when the reception unit 102 received the patient's sample. In addition, information about the day of the week and holidays may be obtained and used from this information. Furthermore, to indicate approximately what time of day the data was collected, it is also effective to express 24 hours as a real number, for example, 13.5 for 1:30:00 PM.

[0023] The reception time 302 may be provided in the reception information 110 by the reception unit 102, which is the information sender, or alternatively, the time of the storage unit 103, which is the receiver, may be used. If there are multiple reception units 102, their clocks may not be synchronized, so it is preferable to use a unified internal clock on the receiver to avoid discrepancies in the sequence of events.

[0024] Furthermore, when the reception unit 102 accepts a sample collection request, the data acquired and stored by the storage unit 103 includes, in addition to patient identification information 301 and reception time 302, some or all of the following: sample type information 303 of the sample to be collected from the patient, reception number 304 indicating the patient's reception order on the system, classification information 305 distinguishing whether the patient is an outpatient or inpatient, the number of patients waiting to be called for sample collection at the time the patient was accepted 306, and requesting source information 310 that instructed the patient to collect a sample. This further improves the accuracy of the prediction unit 101's call time prediction. This data does not necessarily have to be provided directly via reception information 110; for example, it may be provided or referenced from another location (server, system, etc.) using patient identification information 301.

[0025] Specimen type information 303 is information about the type of specimen to be collected. By referring to this field, it is possible to determine, for example, which specimens (serum, whole blood, plasma, etc.) will be collected, or how many types of specimens will be collected.

[0026] Reception number 304 indicates the order in which the sample was received on the day of collection, according to the clinical laboratory information system. If this information is unavailable, or if the numbering is not standardized across multiple systems, the storage unit 103 may internally count the order for that day and store and use it as reception number 304.

[0027] Classification information 305 is one of the patient attribute pieces of information that indicates whether the patient who registered for sample collection was an inpatient or an outpatient. This is effective when there are differences in the calling patterns between inpatients and outpatients.

[0028] The waiting patient count of 306 indicates how many patients were waiting before the patient who registered for sample collection. If this information is unavailable, the memory unit 103 may internally count, store, and use this number. If the exact number cannot be obtained, the prediction accuracy can be sufficiently ensured by substituting a number with an error margin of about 10% or 10 people, so even if there is a slight difference, that number can be used as the waiting patient count of 306.

[0029] Requester information 310 contains information about which department within the medical institution instructed the patient to collect a specimen. This information is expected to be effective when there are differences in calling patterns between departments, such as some departments calling patients earlier than others.

[0030] After receiving the reception information 110, the memory unit 103 then issues a prediction instruction 111 to the prediction unit 101.

[0031] The prediction unit 101 can directly access the information received in the reception information 110 and the information processed based on it, and uses a pre-built machine learning model to predict the call time for sample collection. The prediction unit 101 provides the prediction result 112 to the storage unit 103, and the prediction result 112 is stored in the storage unit 103. The reason for storing it in the storage unit 103 is to later evaluate how much the prediction result deviated from the actual call time and to use that to make improvements.

[0032] The memory unit 103 uses the prediction result information 113 to send the call time prediction result back to the reception unit 102. The time in the prediction result 112 is expressed as a real number, while the time in the prediction result information 113 is expressed as a format such as 1:30:00 PM.

[0033] The reception desk 102 prints the date and time of reception 601, patient name 602, reception number 603, and the estimated call time for sample collection 604 on the sample collection reception form 600 shown in Figure 5, which is used to inform the patient that the reception has been successfully completed. When printing the estimated call time 604, the prediction accuracy 605 is also printed based on past conditions. For example, by including the prediction accuracy 605, such as "There is an 80% probability that the difference between the estimated call time and the actual call time will be within ±4 minutes," the patient can understand how accurate the prediction is.

[0034] Based on this 604 call time prediction information, patients can leave the sample collection waiting room to run other errands and return when their turn comes. This is expected to allow patients to make better use of their waiting time and reduce their stress. Furthermore, by encouraging more patients to leave the waiting room, it is hoped that this will contribute to avoiding the "three Cs" (closed spaces, crowded places, and close-contact settings), which is a common challenge in the current COVID-19 pandemic, and reduce the risk of infection between patients.

[0035] When it is the turn of a patient who has completed the sample collection registration at the reception unit 102, the calling unit 104 displays the patient's registration number on the monitor and prompts the patient to move to the sample collection location. At the same time, the calling unit 104 transmits the call information 114 to the storage unit 103 to notify it that the patient has been called.

[0036] The call information 114 includes at least patient identification information 301 to identify the called patient. Using the patient identification information 301 as a key, it is possible to link it with the reception information 110, calculate the features necessary for building the machine learning model of the prediction unit 101, and store them together in the storage unit 103. At the same time, the storage unit 103 must obtain the call time 308 (Figure 3). This information is based on the time, year, month, day, hour, minute, and second, when the patient was called for sample collection. However, due to the specifications of the call unit 104, it may be the time when the patient travels to the sample collection location and arrives after being called. In this case, although the prediction accuracy will be slightly reduced, it is still sufficient for prediction and can be used as is as the call time 308 for prediction.

[0037] The call time 308 may be provided in the call information 114 by the call unit 104, which is the information sender, or alternatively, the time of the storage unit 103, which is the receiver, may be used. Since the clocks of the reception unit 102 and the call unit 104 may not be synchronized, it is preferable to use the internal clock of the receiver to avoid discrepancies in the sequence of events.

[0038] When the time specified in the start time 501 for machine learning model reconstruction on the parameter setting screen 500 shown in Figure 4, which belongs to the setting unit 107, the evaluation unit 106 constructs and updates the machine learning model using the information stored in the memory unit 103. It is generally desirable to set this to late at night after the end of business operations, when sample collection operations have been completed.

[0039] Furthermore, when constructing the machine learning model, data stored in the memory unit 103 from the day before the target date (today) to the period specified in the training period 502 is used as training data. Typically, this period is specified to be between 14 and 84 days. While longer periods tend to improve prediction accuracy, if the call pattern for sample collection changes due to changes in the sample collection operation, the time it takes for the prediction to adapt to the change becomes longer, and the prediction accuracy decreases during that period. Conversely, shorter periods result in lower prediction accuracy, but the time required for the prediction results to adapt to changes in the call pattern is shorter.

[0040] Furthermore, the setting of the learning period 502 can be automated by checking and enabling the automatic setting button 503. The evaluation unit 106 evaluates the call time predicted by the prediction unit 101 using the method specified in the judgment index 504. For example, if "probability of the error being within 4 minutes" is specified, the evaluation unit 106 calculates the ratio of patients whose error falls within ±4 minutes from the difference between the predicted call time and the actual call time, and determines that a higher ratio is better, i.e., the prediction accuracy is higher. The learning period 502 is automatically changed in 7-day increments of 7 days, 14 days, ... 84 days to build a machine learning model, and on the final day, the setting that provides the best result for the index "probability of the error between the predicted call time and the actual call time being within ±4 minutes" is automatically set as the learning period 502. After that, the machine learning model used by the prediction unit 101 is reconstructed and used for predictions on subsequent days.

[0041] Normally, when a medical institution revises its specimen collection procedures to improve operations, and the call time becomes earlier from a certain day onward, the prediction accuracy temporarily deteriorates because the system learns using data from before the revision. However, when this system is implemented, immediately after the revision, the learning period 502 is automatically shortened, and the proportion of data from after the revision is increased as much as possible, so the period of deterioration can be expected to be shortened. Furthermore, after a certain period has passed since the revision, the learning period 502 is automatically lengthened, which can be expected to improve the prediction accuracy.

[0042] The display unit 105 is a terminal (PC: Personal Computer, mobile terminal, mobile phone, etc.) equipped with a display that shows a monitor screen to inform patients in the waiting room or other area of ​​the calling status of patients waiting for sample collection, as shown in the predicted time guidance monitor screen 700 in Figure 6. The monitor screen displays the current date and time 701, the reception number of the patient waiting for sample collection 702, the calling status of each reception number 703, the reception time of each reception number 704, the predicted calling time of each reception number 705, the actual calling time of each reception number 706, and the prediction accuracy of the call prediction 707, all of which are obtained via the monitor display information 115, and the information is updated as needed.

[0043] The current date and time (701) displays the current time, helping patients waiting to be called to understand how many minutes until they are called, compared to their estimated call time (705). Reception number 702 indicates which patient the information in the horizontal row belongs to.

[0044] Status 703 displays the call status. The display changes depending on the situation: "Waiting for Call" if the sample collection has been accepted but the user has not yet been called; "Called" if the user has already been called for sample collection; and "Calling" if the user has just been called for sample collection. For "Calling" and "Called," it is easy to automatically switch the display after a certain period of time, for example, one minute. However, if information that sample collection has started is available, the display can be switched from "Calling" to "Called" based on that information. The text color and background color of the display can also be changed according to Status 703. Symbols and diagrams can also be used instead of text to make it easier to understand.

[0045] The reception time 704 displays the reception time for sample collection for each reception number. It may be possible to select whether to display only the hours and minutes, or to display the hours, minutes, and seconds together.

[0046] The call prediction time 705 displays the estimated call time for sample collection for each reception number. It may be possible to select whether to display only hours and minutes, or to display hours, minutes, and seconds together. Alternatively, the remaining time (how many minutes or seconds until the call) may be displayed.

[0047] The call time 706 is displayed when a patient waiting for sample collection is called. It may be possible to select whether to display only the hours and minutes, or to display the hours, minutes, and seconds together. This information is more useful to patients who will be called later than to the patient who was called. Patients who were called before them can see and use this information to understand how much later or earlier they were called compared to the call time prediction system 10's predicted call time 705.

[0048] The prediction accuracy 707 is the accuracy of the call prediction time 705 obtained from past situations. For example, it presents to patients waiting for a call how accurate the prediction is, such as "There is an 80% probability that the difference between the predicted call time and the actual call time will be within ±4 minutes," and asks them to understand that the prediction may be wrong. The prediction accuracy 707 displayed here can use the judgment index 504 specified in the parameter setting screen 500 (for example, an error of 4 minutes) and the probability calculated by the evaluation unit 106 (for example, 80%).

[0049] The features of this embodiment can also be summarized as follows.

[0050] As shown in Figure 1, the call time prediction system 10 includes at least a first processor (prediction unit 101). The first processor (prediction unit 101) uses machine learning (artificial intelligence) to predict the time when a patient will be called for sample collection, based on at least one of the following: reception time information (reception time 302) indicating the reception time of the patient to be collected; sample type information 303 indicating the type of sample to be collected from the patient; reception number 304 indicating the patient's reception order; inpatient / outpatient classification information 305 indicating whether the patient is an inpatient or an outpatient; and the number of patients waiting to be called for sample collection at reception time 302 (waiting patient count 306). This allows patients to make effective use of their waiting time until sample collection.

[0051] More specifically, as shown in Figures 2 and 3, the call time prediction system 10 includes a storage unit 103 (Figure 1) that stores at least one of the following: reception time information (reception time 302), specimen type information 303, reception number 304, inpatient / outpatient classification information 305, number of patients waiting to be called (waiting patient count 306), and the measured time when a patient is called (call time 308). The first processor (prediction unit 101) performs machine learning to predict the time when a patient will be called for specimen collection based on the data stored in the storage unit 103. This allows the machine learning model to be reconstructed by comparing the predicted and measured times when a patient will be called for specimen collection. The storage unit 103 is composed of a storage device such as memory or an HDD (Hard Disk Drive).

[0052] Specifically, the first processor (evaluation unit 106, Figure 1) changes the machine learning training period 502 (Figure 4) multiple times, constructs a provisional machine learning model for each training period, and determines the training period for the provisional machine learning model that has the highest probability of the difference between the predicted and actual times when a patient is called being within a threshold, which is an indicator of prediction accuracy (judgment indicator 504, Figure 4). Then, after setting the determined training period, the first processor (evaluation unit 106, Figure 1) reconstructs the machine learning model used in the business. This makes it possible to automatically set the training period for machine learning and reconstruct the machine learning model.

[0053] The call time prediction system 10 includes output devices (printer in the reception unit 102, display in the display unit 105, Figure 1). The first processor (evaluation unit 106, Figure 1) calculates the prediction accuracy, which indicates the probability that the difference between the predicted and actual call times for patients is within a threshold. The output devices output the predicted call times for patients (predicted call time 604 in Figure 5, predicted call time 705 in Figure 6), the thresholds which are indicators of prediction accuracy (4 minutes for prediction accuracy 605 in Figure 5, 4 minutes for prediction accuracy 707 in Figure 6), and the prediction accuracy (80% for prediction accuracy 605 in Figure 5, 80% for prediction accuracy 707 in Figure 6).

[0054] This allows patients to check the predicted time they will be called for sample collection and its accuracy. In this embodiment, the output device is a printer or a display. This allows patients to visually confirm the predicted time they will be called and its accuracy.

[0055] In this embodiment, the first processor (prediction unit 101, Figure 1) uses machine learning to predict the time when a patient will be called for sample collection, based on data from the storage unit 103, which includes at least the number of patients waiting to be called (number of waiting patients 306, Figure 2). More specifically, the first processor (prediction unit 101, Figure 1) uses machine learning to predict the time when a patient will be called for sample collection, based on data from the storage unit 103, which includes at least the number of patients waiting to be called (number of waiting patients 306, Figure 2) and reception time information (reception time 302, Figure 2). More specifically, the first processor (prediction unit 101, Figure 1) uses machine learning to predict the time when a patient will be called for sample collection, based on data from the storage unit 103, which includes at least the number of patients waiting to be called (number of waiting patients 306, Figure 2), reception time information (reception time 302, Figure 2), and reception number 304. According to the inventor's findings, the number of patients waiting to be called (waiting patient count 306), reception time information (reception time 302), and reception number 304 have the greatest impact on prediction accuracy, in that order.

[0056] The call time prediction system 10 includes a setting unit 107 (Figure 1) for setting the machine learning training period. The setting unit 107 consists of, for example, an input device (keyboard, mouse, etc.) and a display. The first processor displays a parameter setting screen 500 (Figure 4) on the display and accepts input values ​​for each input item on the parameter setting screen 500 via the input device. This makes it easy to set the machine learning training period.

[0057] The call time prediction system 10 includes a reception unit 102 (Figure 1) for registering patients to collect specimens. The reception unit 102 consists of, for example, an input device (such as a touch sensor on a touch panel), a display (such as a touch panel display), and a printer for printing specimen collection registration forms 600. This allows patients to register for specimen collection themselves.

[0058] The call time prediction system 10 includes a call unit 104 (Figure 1) that calls patients to collect specimens. The call unit 104 is, for example, composed of a display. This allows the patient to be notified when it is their turn to collect a specimen.

[0059] As described above, this embodiment makes it possible to improve the accuracy of predicting the time when a patient will be called in for sample collection.

[0060] (Second embodiment) Another embodiment of the present invention will be described below with reference to Figure 7.

[0061] Figure 7 is a schematic diagram of a sample collection call time prediction system 10, which is a second embodiment of the present invention, and is designed with the intention of continuing to use a machine learning model that has been constructed once.

[0062] The call time prediction system 10 consists of a reception unit 102 that handles the registration of patients for sample collection, and a prediction unit 101 that uses machine learning to predict the time from sample collection registration to call.

[0063] The specimen collection call time prediction system 10 of the present invention, in cooperation with or as part of the specimen collection support system 20, provides patients with a predicted call time for specimen collection. To present That's what we're anticipating.

[0064] When linked with the specimen collection support system 20, the reception unit 102 is located on the specimen collection support system 20 side, and the prediction unit 101 is located on the call time prediction system 10 side. The specimen collection support system 20 and the call time prediction system 10 are connected via the medical facility's LAN or the internet and exchange information with each other.

[0065] Patients who have received instructions for specimen collection move to the specimen collection location and operate the reception unit 102 to register themselves in the specimen collection queue. In some medical facilities, the retrieval of medical records is linked to the specimen collection registration; in this case, the reception unit 102 automatically registers the patient in the specimen collection queue based on instructions from the medical record system.

[0066] In conjunction with registration to the waiting list, the reception unit 102 transmits reception information 110 to the prediction unit 101, providing information related to the registered patient.

[0067] As shown in Figure 2, the prediction unit 101 obtains data related to the reception unit 102 through the reception information 110. The reception information 110 includes at least patient identification information 301 for identifying patients who have submitted samples for collection. Patient identification information 301 should preferably use a dual key, either the patient ID and the date, or the reception number and the date.

[0068] One piece of information that the prediction unit 101 must obtain is the sample collection reception time 302. The sample collection reception time 302 includes the year, month, day, hour, minute, and second information of when the reception unit 102 received the patient. In addition, information on the day of the week and holidays may be obtained and used from this information. Furthermore, to indicate approximately what time the data was collected on that day, it is also effective to express 24 hours as a real number, for example, 13.5 for 1:30:00 PM. The reception time 302 may be provided in the reception information 110 by the reception unit 102, which is the information sender, or the time of the prediction unit 101, which is the receiver, may be used instead. If there are multiple reception units 102, their clocks may not be synchronized, so it is preferable to use the internal clock of the receiver to avoid discrepancies in the sequence of events.

[0069] Furthermore, when the reception unit 102 accepts a sample collection request, the prediction unit 101's prediction accuracy of the call time will be further improved if the data obtained by the prediction unit 101 includes, in addition to patient identification information 301 and reception time 302, some or all of the following: sample type information 303 of the sample to be collected from the patient, reception number 304 indicating the patient's reception order on the system, classification information 305 distinguishing whether the patient is an outpatient or inpatient, the number of patients waiting to be called for sample collection at the time the patient was accepted 306, and the requesting party information 310 that instructed the patient to collect a sample. This data does not necessarily have to be provided directly via reception information 110; for example, it may be provided or referenced from another location using patient identification information 301.

[0070] Specimen type information 303 is information about the type of specimen to be collected. By referring to this field, it is possible to determine, for example, whether to collect a serum specimen, a whole blood specimen, or a plasma specimen.

[0071] Reception number 304 is information indicating the order in which the sample was received on the day of collection, according to the clinical laboratory information system. If this information is unavailable, or if the numbering is not standardized across multiple systems, the prediction unit 101 may internally count the order for that day and store and use it as reception number 304.

[0072] Classification information 305 is one of the patient attribute pieces of information that indicates whether the patient who registered for sample collection was an inpatient or an outpatient. This is effective when there are differences in the calling patterns between inpatients and outpatients.

[0073] The waiting patient count of 306 indicates how many patients were waiting before the patient who registered for sample collection. If this information is unavailable, the memory unit 103 may internally count, store, and use this number. If the exact number cannot be obtained, the prediction accuracy can be sufficiently ensured by substituting a number with an error margin of about 10% or 10 people, so even if there is a slight difference, that number can be used as the waiting patient count of 306.

[0074] Requester information 310 contains information about which department within the medical institution instructed the patient to collect a specimen. This information is expected to be effective when there are differences in calling patterns between departments, such as some departments calling patients earlier than others.

[0075] The prediction unit 101 uses the information received in the reception information 110 and the information processed based on it to predict the call time for sample collection using a pre-built machine learning model. The prediction result is transmitted to the reception unit 102 via the prediction result information 113.

[0076] The reception unit 102 prints the reception date and time 601, patient name 602, reception number 603, and the estimated call time for sample collection 604 on the sample collection reception form 600 shown in Figure 5, which is used to inform the patient that the reception has been successfully completed. When printing the estimated call time, the prediction accuracy 605 is also printed based on past conditions. For example, by including the prediction accuracy, such as "There is an 80% probability that the difference between the estimated call time and the actual call time will be within ±4 minutes," the patient can understand how accurate the prediction is.

[0077] Based on this predicted call time information, patients can leave the sample collection waiting room to run other errands and return when it's time. This is expected to allow patients to make better use of their waiting time and reduce their stress. Furthermore, by encouraging more patients to leave the waiting room, it is hoped that this will contribute to avoiding the "three Cs" (closed spaces, crowded places, and close-contact settings), which is a common challenge in the current COVID-19 pandemic, and reduce the risk of infection between patients.

[0078] (Third embodiment) Another embodiment of the present invention will be described below with reference to Figure 8.

[0079] Figure 8 shows an example of applying the sample collection call time prediction system 10 of Figure 1, which is a third embodiment of the present invention, to the management of the number of patients entering and leaving a sample collection waiting room.

[0080] The reception desk 102 is located outside the specimen collection waiting room. Patients who have completed the specimen collection registration process receive a specimen collection registration slip 600 (Figure 5) to find out when they will be called for specimen collection and when they should come to the specimen collection room.

[0081] An entrance gate 803 is installed at the entrance to the specimen collection waiting room, and only patients who meet certain conditions and their accompanying persons are allowed to enter. Therefore, patients who register for specimen collection at the reception desk 102 will spend time in a separate area from the blood collection waiting room after registration.

[0082] A patient identification unit 802 is installed in front of the entrance gate 803. If the patient identification identifier, such as the reception number and patient ID, is printed on the sample collection reception form 600 using a barcode, the patient identification unit 802 can automatically identify the patient using a barcode reader.

[0083] The access control unit 801 accesses the storage unit 103 using patient identification information 301 obtained from the patient identification unit 802 as a key to obtain the estimated call time for sample collection for the patient. If it is within a certain time frame of the estimated call time, for example, 4 minutes before the estimated call time, it permits entry, opens the access gate 803, and prompts the patient to enter the blood collection waiting room. If it is not within the specified time frame (in this case, 4 minutes before the estimated call time), it prompts the patient to try entering again when the time comes.

[0084] Furthermore, as shown in Figure 6, by displaying a message such as "Waiting room entry permitted" in the column for the status of the receiving patient 703 on the predicted time guidance monitor screen 700, it becomes possible to guide patients on which patients are allowed to enter the waiting room. In this case, it is preferable to install the guidance monitor screen 700 outside the waiting room.

[0085] The features of this embodiment can also be summarized as follows.

[0086] As shown in Figure 8, the call time prediction system 10 includes a sensor (patient identification unit 802) that detects patient identification information 301 indicating information that identifies a patient, a gate (entry gate 803) installed in the waiting room, and a second processor (entry management unit 801). The second processor (entry management unit 801) determines whether the patient can enter the waiting room based on the predicted time when the patient corresponding to the patient identification information 301 will be called and the current time. If it is determined that entry is permitted, it opens the gate; if it is determined that entry is not permitted, it closes the gate. This helps to suppress congestion in the waiting room.

[0087] By using the above method, entry into the waiting room for patients awaiting sample collection will be limited to patients and their companions within a certain time frame from the predicted call time, thereby alleviating congestion in the waiting room and reducing the risk of infection from the recent COVID-19 pandemic.

[0088] It should be noted that the present invention is not limited to the embodiments described above, and various modifications are included. For example, the embodiments described above are described in detail to make the present invention easier to understand, and are not necessarily limited to those having all the configurations described. Furthermore, it is possible to replace parts of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add configurations from other embodiments to the configuration of one embodiment. In addition, it is possible to add, delete, or replace parts of the configuration of each embodiment with other configurations.

[0089] Furthermore, some or all of the above configurations and functions may be implemented in hardware, for example, by designing them as integrated circuits. Alternatively, the above configurations and functions may be implemented in software by having the processor interpret and execute programs that implement each function. Information such as programs, tables, and files that implement each function can be stored in memory, a recording device such as a hard disk or SSD (Solid State Drive), or a recording medium such as an IC card, SD card, or DVD.

[0090] For example, the prediction unit 101 or the access control unit 801 may be configured as an integrated circuit. This improves processing speed compared to having the processor execute software to perform the processing. In the third embodiment, the second processor (access control unit 801) is separate from the first processor (prediction unit 101), but they may be configured as a single unit.

[0091] In the above embodiment, the prediction unit 101, storage unit 103, evaluation unit 106, and setting unit 107 are implemented as functions of one server (Figure 1), but they may also be implemented as functions of multiple servers.

[0092] Furthermore, embodiments of the present invention may also be as follows.

[0093] (1) A sample collection call time prediction system 10 comprising a reception unit 102 for patients waiting for sample collection and a prediction unit 101 that predicts the time when a patient will be called for sample collection using machine learning, wherein the prediction unit 101 predicts the time when a patient will be called for sample collection based on one or more of the following: reception time information (reception time 302) when the patient was accepted for sample collection, sample type information 303 for the sample to be collected from the patient, reception number 304 indicating the patient's reception order on the system, inpatient / outpatient classification information 305 that distinguishes whether the patient is an inpatient or an outpatient, and the number of patients waiting for sample collection at the time the patient was accepted (number of waiting patients 306).

[0094] (2) A sample collection call time prediction system 10 of (1), comprising a call unit 104 for calling a patient for sample collection, and a storage unit 103 for storing information related to the patient, wherein the storage unit 103 stores one or more of the following: reception time information (reception time 302) for when the patient was accepted for sample collection, sample type information 303 for the sample to be collected from the patient, reception number 304 indicating the reception order of the patient on the system, inpatient / outpatient classification information 305 for distinguishing whether the patient is an inpatient or an outpatient, the number of patients waiting to be called for sample collection at the time the patient was accepted (waiting patient count 306), and call time information (call time 308) for when the patient was called, and the prediction unit 101 learns to predict the patient call time based on the information stored in the storage unit 103.

[0095] (3)(2) Sample collection call time prediction system 10 The sample collection call time prediction system comprises a setting unit 107 that defines (sets) the period of training data (training period 502) used to build a learning model, and an evaluation unit 106 that compares the predicted call time result with the measured time to evaluate the prediction accuracy, wherein before the prediction unit 101 rebuilds the learning model, the evaluation unit 106 automatically switches the training period to build a provisional learning model, compares and evaluates the predicted call time value with the measured result, and automatically sets a training period that improves prediction accuracy in the setting unit 107, so that the prediction unit 101 can automatically set a more appropriate training period when rebuilding the learning model.

[0096] (4).(2) A sample collection call time prediction system 10 comprising a setting unit 107 that defines (sets) an indicator of prediction accuracy (judgment indicator 504), an evaluation unit 106 that calculates prediction accuracy using the defined indicator, and multiple Sample collection A sample collection call time prediction system comprising a display unit 105 that monitors and displays the predicted call status of waiting patients, characterized in that when the reception unit 102 or the display unit 105 presents the predicted call time for sample collection of a patient, it presents together an index of prediction accuracy (judgment index 504) defined in the setting unit 107 and a prediction accuracy 605 calculated by the evaluation unit 106.

[0097] (5).(2) A sample collection call time prediction system 10 comprising: an entry management unit 801 that determines whether a patient who has registered for sample collection is allowed to enter the waiting room; a patient identification unit 802 that obtains the patient's identification information; and an entry gate unit (entry gate 803) positioned to separate the sample collection waiting room from other areas, wherein the entry management unit 801 obtains the patient's predicted sample collection time (predicted call time 705) from the storage unit 103 based on the patient identification information obtained from the patient identification unit 802, determines whether entry is permitted based on that information, and the entry gate unit (entry gate 803) opens the gate if entry is permitted and closes the gate if entry is not permitted, thereby controlling the entry and exit of people to and from the sample collection waiting room.

[0098] According to (1)-(5), it becomes possible to predict with higher accuracy the time from sample collection acceptance to sample collection call. One easily implementable prediction method is to use the average value of the same time period as the prediction value, based on past waiting times until call. When using this method, our evaluation showed that the error between the predicted value and the measured value was within ±3 minutes for 13-67% of all samples. In contrast, when predicting using the present invention, a better result was obtained, at 65-91%. [Explanation of Symbols]

[0099] 10. Call Time Prediction System 20. Sample Collection Support System 101 Prediction Section 102 Reception Department 103 Storage section 104 Calling section 105 Display section 106 Evaluation Department 107 Settings Section 110 Reception Information 111 Prediction Instructions 112 Prediction Results 113 Prediction Results Information 114 Call information 115 Monitor display information 301 Patient Identification Information 302 Reception hours 303 Specimen Type Information 304 Reception Number 305 Inpatient and outpatient classification information 306 patients waiting 308 Ring time 310 Requester Information 500 Parameter setting screen 501 Start time for machine learning model reconstruction 502 Learning period 503 Automatic setting button 504 Judgment index 600 Sample Collection Reception Form 601 Reception date and time 602 Patient Name 603 Reception Number 604 Estimated Call Time 605 Prediction accuracy 700 Prediction Time Information Monitor Screen 701 Current date and time 702 Reception Number 703 Status 704 Reception hours 705 Estimated call time 706 Call history time 707 Prediction accuracy 801 Entry Management Department 802 Patient Identification Unit 803 Entrance Gate

Claims

1. A storage unit that stores reception time information indicating the reception times of multiple patients from whom specimens are to be collected, specimen type information indicating the type of specimen to be collected from the multiple patients, reception number indicating the reception order of the multiple patients, inpatient / outpatient classification information indicating whether the multiple patients are inpatients or outpatients, the number of patients waiting to be called for specimen collection at the reception time, and the measured values ​​of the multiple call times when the multiple patients were called. A machine learning model is constructed using the following as input data: the reception time information for each of the multiple patients stored in the memory unit, the call time for each of the multiple patients, the sample type information indicating the type of sample to be collected from the multiple patients, and the number of patients waiting to be called for sample collection at the reception time; and a prediction unit predicts the time at which the multiple patients will be called for sample collection based on at least the sample type information indicating the type of sample to be collected from the multiple patients and the number of patients waiting to be called for sample collection at the reception time, using the machine learning model. The prediction unit evaluates the time at which the multiple patients predicted by the prediction unit will be called for sample collection, A sample collection call time prediction system comprising: The evaluation unit, The machine learning model is rebuilt at a specified time within a day, and the learning period is set to multiple learning periods with regular intervals within the specified period, and a provisional machine learning model is constructed for each learning period. From among the multiple learning periods, the evaluation unit calculates the difference between the predicted time the patient is called and the measured values ​​of the multiple call times, and determines the learning period of the provisional machine learning model that has the highest probability of the calculated difference being within a threshold, which is an indicator of prediction accuracy. After setting the determined learning period, the machine learning model used in the business is reconstructed. A system for predicting the call time for sample collection, characterized by the following features.

2. A sample collection call time prediction system according to claim 1, Equipped with an output device, The evaluation unit, The prediction accuracy, which indicates the probability that the difference between the predicted and actual time of the patient being called is within the threshold, is calculated. The output device is The predicted time when the patient will be called, the threshold value which is an indicator of the prediction accuracy, and the prediction accuracy are output. A system for predicting the call time for sample collection, characterized by the following features.

3. A sample collection call time prediction system according to claim 1, A patient identification unit that detects patient identification information indicating the aforementioned patient identification information, The gate to be installed in the waiting room, An access control unit determines whether the patient can enter the waiting room based on the predicted time when the patient corresponding to the patient identification information will be called and the current time, and controls the gate to open if it is determined that entry is permitted, and to close the gate if it is determined that entry is not permitted. A system for predicting the call time for sample collection, characterized by comprising the following:

4. A sample collection call time prediction system according to claim 1, The evaluation unit, Based on data including at least the number of patients waiting to be called, the time at which the patient will be called for sample collection is predicted using machine learning. A system for predicting the call time for sample collection, characterized by the following features.

5. A sample collection call time prediction system according to claim 4, The evaluation unit, Based on the data, which includes at least the reception time information, the time when the patient will be called in for sample collection is predicted using machine learning. A system for predicting the call time for sample collection, characterized by the following features.

6. A sample collection call time prediction system according to claim 5, The evaluation unit, Based on the data, including at least the aforementioned reception number, the time when the patient will be called in for sample collection is predicted using machine learning. A system for predicting the call time for sample collection, characterized by the following features.

7. A sample collection call time prediction system according to claim 1, The facility includes a reception unit for receiving patients from whom specimens are to be collected. A system for predicting the call time for sample collection, characterized by the following features.

8. A sample collection call time prediction system according to claim 7, The system includes a calling unit to summon the patient from whom the sample is to be collected. A system for predicting the call time for sample collection, characterized by the following features.

9. A sample collection call time prediction system according to claim 1, The machine learning model includes a setting unit for setting the learning period of the aforementioned machine learning model. A system for predicting the call time for sample collection, characterized by the following features.

10. A sample collection call time prediction system according to claim 2, The output device is It is a printer or a display. A system for predicting the call time for sample collection, characterized by the following features.

11. A storage unit that stores reception time information indicating the reception times of multiple patients from whom specimens are to be collected, specimen type information indicating the type of specimen to be collected from the multiple patients, reception number indicating the reception order of the multiple patients, inpatient / outpatient classification information indicating whether the multiple patients are inpatients or outpatients, the number of patients waiting to be called for specimen collection at the reception time, and the measured values ​​of the multiple call times when the multiple patients were called. A machine learning model is constructed using the following as input data: the reception time information for each of the multiple patients stored in the memory unit, the call time for each of the multiple patients, the sample type information indicating the type of sample to be collected from the multiple patients, and the number of patients waiting to be called for sample collection at the reception time; and a prediction unit predicts the time at which the multiple patients will be called for sample collection based on at least the sample type information indicating the type of sample to be collected from the multiple patients and the number of patients waiting to be called for sample collection at the reception time, using the machine learning model. The prediction unit evaluates the time at which the multiple patients predicted by the prediction unit will be called for sample collection, A method for predicting the call time for sample collection in a sample collection call time prediction system comprising: The machine learning model is rebuilt at a specified time within a day, and the learning period is set to multiple learning periods with regular intervals within the specified period, and a provisional machine learning model is constructed for each learning period. From among the multiple learning periods, the evaluation unit calculates the difference between the predicted time the patient is called and the measured values ​​of the multiple call times, and determines the learning period of the provisional machine learning model that has the highest probability of the calculated difference being within a threshold, which is an indicator of prediction accuracy. After setting the determined learning period, the machine learning model used in the business is reconstructed. A method for predicting the call time for sample collection, characterized by the features described above.

Images