In-hospital stroke prediction device, in-hospital stroke prediction system, in-hospital stroke prediction method, and in-hospital stroke prediction program

Abstract

To provide a device for predicting in-hospital onset of stroke.SOLUTION: An in-hospital stroke prediction device includes: a medical record information acquisition unit for acquiring medical record information of a patient upon admission to a hospital, during hospitalization, or both; a risk information computation unit configured to derive information regarding the in-hospital stroke risk of the patient when the medical record information is entered; and a risk information output unit for outputting information regarding the derived in-hospital stroke risk.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to an in-hospital stroke prediction device, an in-hospital stroke prediction system, an in-hospital stroke prediction method, and an in-hospital stroke prediction program. [Background technology]

[0002] Treatment of stroke, such as cerebral infarction, is a race against time: for example, thrombolytic therapy must be performed within 4.5 hours, and thrombectomy must be performed within 16 hours. However, because strokes occur suddenly, it can be difficult to provide appropriate and timely treatment.

[0003] As a technology for assessing the risk of cerebrovascular disease in advance, a system has been proposed that calculates the risk level of cerebrovascular disease based on personal medical information from a pedometer, blood pressure monitor, acceleration sensor (gait abnormalities), microphone (speech abnormalities), and electrocardiograph (arrhythmia), and determines whether to contact a specified contact person (Patent Document 1).

[0004] A system has been proposed that detects signs of disease based on changes over time in biological information such as heart rate, pulse, blood pressure, respiratory rate, body temperature, blood oxygen status, cardiac potential, brain waves, skin temperature, heart rate intervals, brain wave intervals, and body movement, as well as genetic information (Patent Document 2). [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Patent No. 7298970 [Patent Document 2] Japanese Patent Application Publication No. 2019-048061 Summary of the Invention [Problem to be solved by the invention]

[0006] As mentioned above, systems for predicting cerebrovascular disorders and diseases have been proposed, but there is still a challenge in more effectively predicting the occurrence of hospital-onset stroke.

[0007] According to a nationwide survey of in-hospital cerebral infarction conducted by the inventor, in-hospital cerebral infarction occurred at 508 of 539 acute stroke treatment facilities, resulting in an incidence rate of in-hospital cerebral infarction of 94.2%, with 5,341 cases of in-hospital cerebral infarction occurring annually at medical facilities nationwide. Furthermore, slightly more than 10% of cerebral infarction patients develop their stroke in a hospital. Furthermore, when a cerebral infarction occurs in a hospital, it overlaps with the illness for which the patient was hospitalized, and approximately one-third of patients who develop a cerebral infarction in a hospital die or suffer severe sequelae, leading to a poor prognosis. This is true not only for cerebral infarction, but also for cerebral hemorrhage and subarachnoid hemorrhage.

[0008] Therefore, in clinical practice, predicting the occurrence of in-hospital stroke is an important issue. In addition, because in-hospital stroke can occur as a perioperative complication associated with medical procedures such as catheterization, predicting the occurrence of in-hospital stroke is also an important issue from the perspective of medical safety. Therefore, there is a need to predict the occurrence of in-hospital stroke not only from the perspective of clinical practice but also from the perspective of medical safety. [Means for solving the problem]

[0009] The gist of the present invention is as follows. (1) a medical record information acquisition unit that acquires medical record information of patients at the time of admission to the hospital, during their stay, or both; a risk information calculation unit that calculates information regarding the patient's risk of developing an in-hospital stroke when the medical record information is input; and a risk information output unit that outputs information related to the calculated in-hospital stroke risk An in-hospital stroke prediction device comprising: (2) The prediction device described in (1) above, wherein the medical record information at the time of hospitalization includes age, sex, underlying disease, medical history, whether or not a patient has had a stroke, oral medications, drinking history, smoking history, family history, examination history, blood test results, referral letter, or a combination thereof. (3) The prediction device according to (1) or (2) above, wherein the medical record information during hospitalization includes medical records, nursing records, discharge summaries, test results, test findings, surgery records, prescriptions, requests for medical treatment from other departments during hospitalization, or a combination thereof. (4) The prediction device according to any one of (1) to (3) above, wherein the medical record information is medical record information at the time of hospitalization and during hospitalization. (5) Calculating information about the risk of in-hospital stroke Analyzing the text information of the medical record information by text mining to extract words and phrases; and Calculating information regarding the risk of in-hospital stroke using the extracted words and phrases. The prediction device according to any one of (1) to (4) above, comprising: (6) Calculating information about the risk of in-hospital stroke Analyzing the text information of the medical record information by text mining to extract words and phrases; weighting the extracted phrases; and Using the weighted phrases to calculate information about the risk of in-hospital stroke. The prediction device according to any one of (1) to (4) above, comprising: (7) Calculating information about the risk of in-hospital stroke Analyzing character data included in the medical record information; and Calculating information regarding the risk of in-hospital stroke using the analyzed character data. The prediction device according to any one of (1) to (4) above, comprising: (8) Calculating information about the risk of in-hospital stroke Analyzing the text information of the medical record information by text mining to extract words and phrases; Analyzing character data included in the medical record information; and Calculating information regarding the in-hospital stroke risk using the analyzed character data and the extracted words and phrases. The prediction device according to any one of (1) to (4) above, comprising: (9) A prediction device according to any one of (1) to (8) above, wherein calculating the information regarding the risk of in-hospital stroke includes using a trained calculation model that has undergone machine learning using training data including, as input, the subject's medical record information and, as output, information regarding whether the subject has developed a stroke, so that the information regarding the risk of in-hospital stroke is calculated when the medical record information is input. (10) A prediction device according to any one of (1) to (9) above; a medical record server in which the medical record information is stored; An in-hospital stroke prediction system comprising: (11) Obtaining medical record information of patients at the time of admission or during their stay at the hospital; inputting the medical record information to calculate information regarding the patient's risk of developing an in-hospital stroke; and outputting information relating to the calculated in-hospital stroke risk; A method for predicting in-hospital stroke, comprising: (12) A medical record information acquisition function for acquiring medical record information of patients at the time of admission to a hospital or during their stay; a risk information calculation function that calculates information about the patient's risk of developing an in-hospital stroke when the medical record information is input; and A risk information output function that outputs information related to the calculated in-hospital stroke risk. In-hospital stroke prediction program, including [Effects of the Invention]

[0010] According to the present invention, it is possible to provide an apparatus for predicting the onset of stroke in a hospital. [Brief explanation of the drawings]

[0011] [Figure 1] FIG. 1 is a schematic diagram showing an example of the configuration of an in-hospital stroke prediction system 100 including an in-hospital stroke prediction device (also referred to as the present device) 10. [Figure 2]FIG. 2 is a diagram showing an example of the configuration of the device 10. As shown in FIG. [Figure 3] FIG. 3 is a diagram showing another example of the configuration of the device 10. In FIG. [Figure 4] FIG. 4 is a graph showing the incidence rate of in-hospital cerebral infarction (predicted risk of in-hospital cerebral infarction) by age obtained from medical record information at the time of hospitalization. [Figure 5] Figure 5 is a flowchart of the hospital-onset stroke algorithm. [Figure 6] FIG. 6 is a diagram showing another example of the configuration of the device 10. In FIG. [Figure 7] FIG. 7 is a configuration diagram of an example of a training data collection system when collecting training data TD used in machine learning. [Figure 8] FIG. 8 is a configuration diagram of an example of a machine learning system when machine learning is performed. [Figure 9] FIG. 9 is a schematic diagram showing an example of the calculation model M1. [Figure 10] FIG. 10 is a schematic diagram showing another example of the calculation model M1. [Figure 11] FIG. 11 is a configuration diagram of an example of an in-hospital stroke prediction system when the device 10 is used to predict a patient's risk of in-hospital stroke. DETAILED DESCRIPTION OF THE INVENTION

[0012] The present disclosure relates to an in-hospital stroke prediction device that includes a medical record information acquisition unit that acquires medical record information of a patient at the time of admission to a hospital, during hospitalization, or both, a risk information calculation unit that, when the medical record information is input, calculates information related to the patient's risk of in-hospital stroke, and a risk information output unit that outputs the calculated information related to the risk of in-hospital stroke.

[0013] The device disclosed herein (hereinafter also referred to as the present device) can predict the risk of in-hospital stroke (hereinafter also referred to as the in-hospital stroke risk) using a patient's medical record information at the time of admission to a hospital, during their hospitalization, or both. The in-hospital stroke risk refers to the risk of a patient suffering a stroke while hospitalized. The medical record information acquired by the present device includes items that are examined during the patient's regular medical care at the time of admission, during their hospitalization, or both. This allows for simple and comprehensive prediction of in-hospital stroke for hospitalized patients. An in-hospital stroke refers to a stroke that occurs while the patient is hospitalized. The present device can prevent the occurrence of strokes while in the hospital, enable rapid response when a stroke occurs, and improve the patient's life and functional prognosis. A stroke is a disease caused by a cerebrovascular disorder, including cerebral infarction, cerebral hemorrhage, or subarachnoid hemorrhage.

[0014] Traditionally, all hospital staff have been educated about in-hospital stroke, but as mentioned above, the incidence rate of in-hospital cerebral infarction is low at 0.01%, and the incidence rates of in-hospital cerebral hemorrhage and subarachnoid hemorrhage are even lower, so there was an inability to strike a balance between the effort and cost of education and its effectiveness.

[0015] In contrast, this device can classify patients into those at high risk of in-hospital stroke and those at low risk, allowing for tailored education for healthcare professionals working with high-risk patients and those working with low-risk patients. Healthcare professionals refer to doctors, nurses, and other hospital staff. This allows for a balance between the effort and cost of education and its effectiveness, and enables a more rapid response to in-hospital strokes. For example, healthcare professionals working with high-risk patients can receive focused education, such as mandatory e-learning and distribution of FAST materials, while healthcare professionals working with low-risk patients can receive general education, such as distribution of FAST materials, without requiring e-learning.

[0016] The device is preferably capable of screening patients whose in-hospital stroke incidence rate (predicted in-hospital risk) is equal to or greater than a predetermined value, for example, 0.1% or greater, 0.5% or greater, 1.0% or greater, 1.5% or greater, or 2.0% or greater.

[0017] The device can preferably output items from medical record information at the time of admission, medical record information during hospitalization, or a combination thereof, when the in-hospital stroke incidence rate (predicted in-hospital stroke risk) is, for example, 0.1% or more, 0.5% or more, 1.0% or more, 1.5% or more, 2.0% or more, etc. The device can output the patient's age and items according to the in-hospital stroke incidence rate (predicted in-hospital stroke risk), and for example, when the patient is 57 years old or older, can output items such as a history of stroke, abnormally high white blood cell and urea nitrogen levels, and abnormally low platelet count.

[0018] 1 is a schematic diagram showing an example of the configuration of an in-hospital stroke prediction system 100 (hereinafter also referred to as the present system) including the present device 10. The present device 10 can be connected to a medical record server 30 that stores patient medical record information directly or via a network 20. The network 20 is a wireless or wired communication network such as the Internet, a LAN, or Bluetooth.

[0019] The device 10 and the medical record server 30 are information processing devices such as computers and servers. The device 10 can be a personal computer, tablet, smartphone, etc. The device 10 and the medical record server 30 can also be an electronic medical record terminal equipped with a display.

[0020] The device 10 and the medical record server 30 may be configured as a single information processing device, or may be a collection of multiple physically separate information processing devices. In this case, each of the multiple information processing devices may have the same functions, or may have the functions of a single device 10 in a distributed manner, or may have the functions of a single medical record server 30 in a distributed manner.

[0021] 2 is a diagram showing an example of the configuration of the device 10. The device 10 has a medical record information acquisition unit 101 that has a function of acquiring medical record information in which a patient's medical record is recorded. The device 10 also has a risk information calculation unit 102 that has a calculation function of calculating information related to the patient's in-hospital stroke risk, and a risk information output unit 103 that outputs the calculated information related to the in-hospital stroke risk.

[0022] 3 is a diagram showing another example of the configuration of the device 10. The device 10 may have a medical record information acquisition unit 101, a memory unit 12, and a processing unit 15, and the processing unit 15 may include a receiving and processing unit 151, a risk information calculation unit 102, and a risk information output unit 103. The device 10 may also have a display unit 13 and an operation unit 14.

[0023] The device 10 can acquire medical record information from the medical record server 30 via the medical record information acquisition unit 101. The medical record information acquisition unit 101 can be a communication device and is implemented as hardware, firmware, communication software such as a TCP / IP driver or a PPP driver, or a combination of these. Communication by the communication device can be wireless or wired communication. The acquired medical record information can be stored in the storage unit 12.

[0024] The device 10 can acquire medical record information from the medical record server 30 via a communication device. The communication device may receive medical record information from the medical record server 30 via serial communication using a USB cable. The communication device may also have an interface circuit for short-range wireless communication according to a communication method such as Bluetooth (registered trademark) and may receive radio waves from the medical record server 30. The communication device may also have a receiving circuit for receiving various signals corresponding to data from the medical record server 30 via infrared communication or the like. The communication device may also have a communication interface circuit for a wired LAN.

[0025] The medical record information acquisition unit 101 may include an input / output device that detachably holds a portable storage medium instead of or in addition to the communication device. The input / output device can acquire the medical record information stored in the portable storage medium and store the acquired medical record information in the storage unit 12 of the device.

[0026] The storage unit 12 is, for example, a semiconductor memory device such as a ROM or RAM. The storage unit 12 may be, for example, a magnetic disk, an optical disk, a magneto-optical disk, a nonvolatile semiconductor memory, or any other storage device capable of storing data. The storage unit 12 stores an operating system program, a driver program, an application program, data, etc. used for processing in the processing unit 15. The risk information calculation unit 102 and the risk information output unit 103 may be included in the processing unit 15. The computer programs stored in the storage unit 12 may be downloaded online and installed in the storage unit 12, or may be installed in the storage unit 12 from a computer-readable portable recording medium such as a CD-ROM or DVD-ROM using a known setup program, etc. The device 10 may be connected to a storage unit implemented in the cloud to perform the functions of the storage unit 12.

[0027] The risk information calculation unit 102 transmits the calculation results to the risk information output unit 103. The risk information output unit 103 outputs information on the calculation results. The risk information output unit 103 can output information on the in-hospital onset risk calculated by the risk information calculation unit 102 to a display unit 13 such as a display, or transmit it to another device. The display unit 13 may be a liquid crystal display, an organic EL display, or the like. The display unit 13 may be provided in the device 10, or may be connected to the device 10 directly or via a network. The network may be a wireless or wired communication network such as the Internet, LAN, Bluetooth, or the like.

[0028] The information relating to the risk of in-hospital onset calculated by the risk information calculation unit 102 is preferably an image, a numerical value, a character, a symbol, a sound, or a combination thereof. The information relating to the risk of in-hospital onset is more preferably a binary value of presence or absence or high or low risk of in-hospital onset, a score of the risk of in-hospital onset, or a combination thereof. The score of the risk of in-hospital onset can be a scale of 1 to 5 or the like, or a percentage of 0 to 100% of the risk of in-hospital onset, or the like.

[0029] The operation unit 14 can be a pointing device such as a keyboard, a mouse, and / or a touch panel. A user of the device 10 can operate the device 10 using the operation unit 14. When operated by the user of the device 10, the operation unit 14 generates a signal corresponding to the operation. The generated signal is then supplied to the processing unit 15 as an instruction from the user.

[0030] The processing unit 15 is a processing device that loads the operating system program, driver program, application program, control program, etc. stored in the storage unit 12 into memory and executes instructions included in the loaded programs. The processing unit 15 is, for example, an electronic circuit such as a CPU (Central Processing Unit), MPU (Micro Processing Unit), DSP (Digital Signal Processor), or GPU (Graphics Processing Unit), or a combination of various electronic circuits.

[0031] Processing unit 15 may be realized by integrated circuits such as ASICs (Application Specific Integrated Circuits), PLDs (Programmable Logic Devices), FPGAs (Field Programmable Gate Arrays), MCUs (Micro Controller Units), etc. Processing unit 15 may be a single component or a collection of multiple physically separate processors. For example, processing unit 15 may be implemented with multiple processors operating cooperatively in parallel to execute instructions.

[0032] The processing unit 15 can function as a receiving processing unit 151, a risk information calculation unit 102, and a risk information output unit 103 by executing various commands contained in an application program (control program) stored in the memory unit 12.

[0033] The receiving processing unit 151 can receive the medical record information stored in the medical record server 30 via the medical record information acquiring unit 101 and store the received medical record information in the storage unit 12.

[0034] The medical record information, which is input data, is received and processed by the receiving processing unit 151. The risk information calculation unit 102 uses the medical record information stored in the storage unit 12 as input data, calculates information regarding the patient's risk of in-hospital onset, and stores the information in the storage unit 12.

[0035] The medical record information acquired by the medical record information acquisition unit 101 includes items recorded and examined during normal admission and medical treatment of patients at the time of admission, during hospitalization, or both, and does not include items resulting from special tests. The medical record information can be image data including DPC data, characters, numbers, etc., or text data including characters, numbers, etc. The medical record information can be acquired by the medical record information acquisition unit 101 in a state linked to the patient ID, the date of recording, etc. DPC data is collected from acute care hospitals and other facilities throughout Japan, and includes information related to medical expense claims for medications, surgeries, etc. (E File, F File, D File, etc.), and Form 1, etc., which includes detailed information on patient attributes, pathological conditions, etc.

[0036] The medical record information may be inputted into the medical record information acquisition unit 101 through the operation unit 14 of the device 10 or the operation unit of the medical record server 30, and the medical record information may be acquired by the medical record information acquisition unit 101. The medical record information inputted through the operation unit 14 of the device 10 or the operation unit of the medical record server 30 may be inputted as letters, numbers, or a combination thereof into an input box for each item of the medical record information displayed on the display unit 13, or letters, numbers, or a combination thereof may be inputted as text information in a free format.

[0037] The processing unit 15 may acquire the medical record information recorded in a predetermined format in the medical record server 30 via the medical record information acquisition unit 101. For example, the processing unit 15 can acquire the medical record information structured as variables in the electronic medical chart or DPC data by issuing an instruction to acquire the medical record information to the processing unit 15 via the operation unit 14.

[0038] For example, age, sex, and primary disease at the time of admission may be obtained from the data on the illness or injury that led to admission in DPC Data Form 1, age may be calculated from the date of birth and date of admission, the presence or absence of each medical history may be obtained from the diagnosis name data in the electronic medical record or the complication data in DPC Data Form 1, the presence or absence of abnormal high or low values ​​in each test value may be obtained from the abnormality determination data of the test results in the electronic medical record, and the presence or absence of antithrombotic therapy may be obtained from the treatment order data in the electronic medical record.

[0039] The acquired medical record information may be stored in the storage unit 12, and may be converted by the receiving and processing unit 151 into a predetermined data format that allows the risk information calculation unit 102 to calculate information related to the risk of in-hospital onset.

[0040] The medical record information at the time of admission is the patient's initial information upon admission. The medical record information at the time of admission preferably includes age, sex, underlying disease, medical history, whether or not a patient has had a stroke, oral medications, alcohol history, smoking history, family history, examination history, blood test results, a referral letter, or a combination thereof. The medical history includes whether or not the patient has had diabetes, hypertension, dyslipidemia, stroke, atrial fibrillation, etc., and may be recorded as the patient's medical record information, for example, up to one year prior. The referral letter may also include a description of the medical history. The blood test results may include whether or not each test value has abnormally high or low values. Examples of blood test results include test results for BNP, PT, APTT, creatinine, urea nitrogen (BUN), white blood cell count (WBC), red blood cell count, hematocrit, hemoglobin, and platelet count (PLT). Information regarding the risk of in-hospital onset may be calculated using the same weighting for each item of the medical record information at the time of admission, or may be combined with different weightings to calculate information regarding the risk of in-hospital onset.

[0041] The hospitalization medical record information is longitudinal information of the patient during hospitalization. The hospitalization medical record information is preferably a medical record, a nursing record, a discharge summary, test results, test findings, a surgical record, a prescription, requests for medical treatment from other departments during hospitalization, or a combination thereof. Each item of the hospitalization medical record information may be weighted equally to calculate information related to the risk of in-hospital onset, or may be weighted differently and combined to calculate information related to the risk of in-hospital onset.

[0042] A medical record is a chart in which a doctor records and preserves a patient's medical treatment, progress, etc. The medical record can be an electronic medical record. Medical records may include discharge summaries, test results, test findings, test images, surgery records, prescriptions, referral letters, etc. A nursing record is a document that records and preserves a nurse's thoughts regarding the purpose and necessity of nursing care, the care provided, etc. A discharge summary is a written summary summarizing an inpatient's medical history, physical examination findings at the time of admission, test results, medical treatment received during hospitalization, etc. It may include a discharge summary when a patient is discharged from another hospital and admitted to this hospital, or a discharge summary when the patient has been admitted and discharged from this hospital at least once. For example, if a patient has been admitted and discharged from this hospital within the past year, the discharge summary may include the discharge summary at that time. Test results are documents showing the results of various tests, such as blood tests and urine tests. Test results preferably include vital sign monitoring results, blood test results, imaging test results, or a combination thereof.

[0043] Laboratory findings are the doctor's findings based on the results of various tests. Surgical records are records of intraoperative findings and surgical procedures in surgical cases. A prescription is a document that lists the type, amount, and dosage of medication needed to treat a patient's illness. A referral letter, also known as a medical information document, is a document that conveys patient information from a family doctor to the referring department or medical institution.

[0044] The medical record information acquisition unit 101 acquires medical record information at the time of admission of the patient, medical record information during hospitalization, or a combination thereof, and preferably acquires medical record information at the time of admission of the patient and medical record information during hospitalization.

[0045] Medical record information at the time of admission can be used to screen patients at high and low risk of in-hospital onset early in the course of admission. Medical record information during hospitalization can be used to predict which hospitalized patients are at high risk of in-hospital onset. Therefore, by acquiring a patient's medical record information at the time of admission and during hospitalization and calculating information about the risk of in-hospital onset based on both, more accurate predictions are possible. In this case, the risk information calculation unit 102 may calculate information about the risk of in-hospital onset using the same weighting for the medical record information at the time of admission and the medical record information during hospitalization, or may weight the analysis results of either and combine them to calculate information about the risk of in-hospital onset.

[0046] When the medical record information stored in the medical record server 30 for each patient is acquired by the medical record information acquisition unit 101 of the device 10, the risk information calculation unit 102 calculates information related to the patient's risk of in-hospital onset. For example, if the medical record information at the time of admission includes age, history of stroke, abnormally high white blood cell and urea nitrogen values, and abnormally low hematocrit and platelet count values, and this information is input into the risk information calculation unit 102, it can calculate that the risk of in-hospital onset is 1%.

[0047] Preferably, the risk information calculation unit 102 analyzes character data included in the medical record information and calculates information related to the risk of in-hospital onset using the analyzed character data. The characters can be numbers, alphabets, hiragana, katakana, kanji, symbols, etc., and are preferably numbers, alphabets, or a combination thereof. For example, the character data may be alphabetic data indicating an abnormality, such as H or L, or numerical data of a test value, or data combining a number and a symbol indicating a range, such as <50 or <25.

[0048] For example, when calculating the risk of in-hospital onset as a binary value of high or low, the risk information calculation unit 102 analyzes character data such as age, gender, medical history, blood test findings, and vital signs, and calculates that the risk of in-hospital onset is high if the character data is related to the risk of in-hospital onset, and calculates that the risk of in-hospital onset is low if the characteristics of the character data are not related to the risk of in-hospital onset.

[0049] Furthermore, for example, when calculating the risk of in-hospital onset as a scale of 1 to 5 or a percentage of 0 to 100%, the risk information calculation unit 102 can analyze character data such as age, sex, medical history, blood test findings, and vital signs, and calculate the scale or percentage of the risk of in-hospital onset based on whether or not this character data contains character data related to a high risk of in-hospital onset.

[0050] The risk information calculation unit 102 can calculate the in-hospital stroke occurrence probability (in-hospital stroke occurrence risk) using logistic regression analysis. Logistic regression analysis is a statistical method that can explain and predict the probability of a "binary outcome (objective variable)" occurring from several factors (explanatory variables). When the occurrence probability is p, the logit (= log(p / (1-p))) can be estimated linearly using each factor (medical record information), which is character data. For example, β0, β1, β2, β3, ... can be estimated using a linear model formula such as logit = intercept β0 + estimated value for age β1 × age + estimated value for gender β2 × gender + estimated value for history of diabetes β3 × presence or absence of history of diabetes + ...

[0051] For example, if we create data rules for character data, such that gender is 1 if male and 0 if female, and whether or not a person has a history of diabetes is 1 if they have, and 0 if they have not, then for estimation, if a 60-year-old man has a history of diabetes, the logit is estimated as logit = β0 + β1 × 60 + β2 + β3, and if a 50-year-old woman has no history of diabetes, the logit is estimated as logit = β0 + β1 × 50.

[0052] By converting this logit back into the probability of occurrence p, the probability of occurrence can be estimated, and by calculating p = EXP(logit) / (1+EXP(logit)), the probability of occurrence of in-hospital stroke (in-hospital occurrence risk) can be calculated.

[0053] The medical record information for each factor, which is character data, may include, for example, the underlying disease at the time of admission, age, gender, history of each condition (diabetes, hypertension, dyslipidemia, stroke, atrial fibrillation, etc.), smoking status, presence or absence of abnormally high or low values ​​for each test value of BNP, PT, APTT, creatinine, BUN, white blood cell count, red blood cell count, hematocrit, hemoglobin, and platelet count, and presence or absence of antithrombotic therapy. The estimated values ​​can be estimated using the medical record information for each of these factors. For each estimated value β of each of the above factors determined to be used for estimating the estimated values, factors for which the null hypothesis is rejected using the STAPWISE method with a null hypothesis of β = 0 and a significance level of 5% on both sides can be selected.

[0054] β = 0 represents a state in which the factor does not contribute to predicting the probability of stroke occurrence, and is the probability of the estimated β occurring when the true β is assumed to be 0. A two-sided 5% significance level means that if the probability of the estimated β occurring is less than 5% when the true β is assumed to be 0, then such a rare event does not occur, and the assumption of β = 0 is rejected, and it is determined that the factor contributes to predicting the probability of stroke occurrence. Furthermore, if more data is obtained and more factors are used, the number of factors adopted will also increase, and even if the same factors are used as data increases, the accuracy of the estimation will increase as the data increases, making it easier to say that they can be rejected at a two-sided 5% significance level, so the number of factors will increase.

[0055] The discarded factors include, for example, whether the patient has a history of cerebral infarction, whether or not the patient has an abnormally low BUN value, whether or not the patient has an abnormally high BUN value, whether or not the patient has an abnormally low WBC value, whether or not the patient has an abnormally high WBC value, whether or not the patient has an abnormally low PLT value, whether or not the patient has an abnormally high PLT value, and age. The probability of in-hospital cerebral infarction (risk of in-hospital cerebral infarction) can be calculated based on the discarded factors. The risk information calculation unit 102 can calculate the probability of in-hospital cerebral infarction (risk of in-hospital cerebral infarction) of the patient based on the presence or absence of each item represented by two bits, 0 or 1, and multiplication of the age and the estimated value of each item. The risk information calculation unit 102 calculates the sum of the presence or absence of each item represented by two bits, 0 or 1, multiplied by the age and the estimated value of each item, using the logit x as the sum: Probability of in-hospital stroke (in-hospital stroke risk) (%) = e (x) / (1+e (x) ) (1) It can be calculated as follows.

[0056] The risk information calculation unit 102 can also calculate information relating to the risk of in-hospital onset by extracting words and phrases (hereinafter also referred to as keywords) relating to the risk of in-hospital onset from text included in the medical record information.

[0057] For example, when calculating the risk of in-hospital onset as a binary value of high or low, the risk information calculation unit 102 can investigate keywords extracted from the medical record information and calculate the risk of in-hospital onset as high if the keywords related to a high risk of in-hospital onset are included, and calculate the risk of in-hospital onset as low if the keywords related to a high risk of in-hospital onset are not included.

[0058] For example, when calculating the risk of in-hospital onset as a scale of 1 to 5 or as a percentage of 0 to 100%, the risk information calculation unit 102 can calculate the scale or percentage of in-hospital onset risk based on whether or not keywords related to a high risk of in-hospital onset are included, the type of keywords, the frequency of keyword appearance, etc.

[0059] The risk information calculation unit 102 can analyze the text data of the medical record information and output information related to the risk of in-hospital onset. The analysis of the text data includes extracting keywords for outputting information related to the risk of in-hospital onset using keywords contained in the text and using the extracted keywords. The risk information calculation unit 102 extracts keywords contained in the text and calculates information related to the risk of in-hospital onset using the extracted keywords. The extraction of keywords can be performed using methods used in text processing, such as keyword search or morphological analysis, which are mechanical processing of text.

[0060] Calculating information related to the risk of in-hospital onset using the extracted keywords includes, for example, investigating the keywords themselves, the location where the keywords appear, the frequency with which the keywords appear, the proximity of the keywords to other keywords, etc., and calculating information related to the risk of in-hospital onset based on the results of the investigation. For the investigation, methods used in language processing can be adopted, such as classification of searched keywords, statistical methods, text mining, machine learning, etc.

[0061] Preferably, calculating information related to the risk of in-hospital onset includes analyzing text information in the medical record information by text mining to extract keywords, and using the extracted keywords to calculate information related to the risk of in-hospital onset. The extracted keywords may be weighted equally to calculate information related to the risk of in-hospital onset, or may be weighted differently and combined to calculate information related to the risk of in-hospital onset. For example, keywords can be extracted by analyzing physician findings (SOAP) recorded in the electronic medical record by text mining.

[0062] The text contained in medical records includes, for example, the item names in the entry fields. Therefore, if the extracted keywords are the item names and their contents, the results of an investigation into the proximity of the two keywords can provide a more precise understanding of the contents, allowing for a more accurate calculation of the risk of in-hospital onset.

[0063] Calculation of information relating to the risk of in-hospital stroke may be performed by combining analysis of the character data contained in the medical record information with analysis using text mining to extract words and phrases, or by combining image analysis and / or machine learning, as described below.

[0064] When the medical record information includes images, the risk information calculation unit 102 calculates information related to the risk of in-hospital onset using the analysis results of the images. The images included in the medical record are, for example, images obtained by CT, MRI, X-ray examination, ultrasound examination, endoscopic examination, etc.

[0065] The image analysis results may include data contained in the image, test findings on the image, etc. The risk information calculation unit 102 can calculate information regarding the risk of in-hospital onset based on the analysis results of the input image. As an image analysis method, a method used for image analysis of medical images can be adopted, for example, by extracting characteristic parts of the image, obtaining color and / or pattern, and obtaining an analysis result regarding the presence or absence of a suspected lesion. Furthermore, these can be performed using pattern matching by machine learning, data mining, etc., to obtain an analysis result regarding the presence or absence of a suspected lesion. The medical image used for image analysis can be one image or two or more images.

[0066] For example, when a CT medical image of the brain is created and a radiologist interprets the image and finds no abnormalities in the cerebral blood vessels, image analysis may reveal an analysis result indicating the presence of a lesion related to the risk of stroke. For example, if image analysis reveals the presence of a lesion related to the risk of cerebral infarction, the medical record may include the term "suspected cerebral infarction" as an analysis result of the image. When this medical record is input, the risk information calculation unit 102 can calculate a high risk of in-hospital onset. Therefore, even if the radiologist overlooks an abnormality in the cerebral blood vessels in the medical image, the output indicating a high risk of in-hospital onset allows for prompt response by the attending physician or specialist.

[0067] (Machine Learning) Preferably, calculating information about the risk of in-hospital onset includes using a trained calculation model that has been subjected to machine learning using training data that includes, as input, the subject's medical record information and, as output, information about whether the subject has had a stroke, so that information about the risk of in-hospital onset is calculated when the medical record information is input. A learning model can be created by performing machine learning on the medical record information at the time of admission and the medical record information during admission for groups with and without in-hospital stroke. The created learning model can be used to output information about the risk of in-hospital onset.

[0068] 6 is a diagram showing an example of the configuration of the present device 10 when the risk information calculation unit calculates information about the risk of in-hospital cerebral infarction or stroke using machine learning. The present device 10 may include a risk information calculation unit having a learning function for training a calculation model and a calculation function for calculating information about the patient's risk of in-hospital cerebral infarction or stroke using the trained calculation model. Data such as a calculation model M1 and training data TD, which will be described later, may be stored in the storage unit 12.

[0069] The processing unit 15 can function as a receiving processing unit 151, a conversion processing unit 152, a learning processing unit 153, a risk information calculation unit 102, and a risk information output unit 103 by executing various commands included in an application program (control program) stored in the storage unit 12. Examples of the functions of the receiving processing unit 151, the conversion processing unit 152, the learning processing unit 153, the risk information calculation unit 102, and the risk information output unit 103 will be described with reference to Figures 7 to 11.

[0070] Figure 7 shows a configuration diagram of an example of a training data collection system for collecting training data TD used in machine learning. Medical record information of the subject is obtained from the medical record server 30. The presence or absence of a stroke in the subject is also obtained. The subjects are divided into two groups, those who have developed a stroke and those who have not, and deep learning is performed to create a learning model. A patient is someone whose risk of in-hospital onset is predicted using this device 10, and a subject is someone from whom training data is obtained in this device 10.

[0071] Data 121 of the medical record information of the measured subject and data 125 of whether the subject has had a stroke can be stored in the storage unit 12. The stored data can be displayed on the display unit 13.

[0072] As described above, the medical record information data 121 and the stroke occurrence data 125 are collected as training data TD. The stroke occurrence data 125 may be a binary value, with 0 representing the absence of stroke in the subject and 1 representing the presence of stroke, or vice versa.

[0073] The learning processing unit 153 generates a trained calculation model M1 trained using the teacher data TD stored in the storage unit 12, and stores the generated trained calculation model M1 in the storage unit 12.

[0074] A configuration diagram of an example of a machine learning system when machine learning is performed is shown in Figure 8. During machine learning, data 121 of measured medical record information and data 125 of the presence or absence of stroke stored in the storage unit 12 are used as training data TD.

[0075] A learning device composed of a conversion processing unit 152 and a learning processing unit 153 preprocesses medical record information data 121 as desired, and trains a calculation model using training data TD having the preprocessed medical record information data as input data and stroke occurrence / presence data 125 as output labels. A calculation model M1 can be generated by updating the weighting variables of the learning model so that the error in information regarding in-hospital onset risk calculated (estimated) based on the input medical record information data 121 is reduced or matched with the stroke occurrence / presence data 125, which is the output label. Deep learning is used to generate the calculation model M1 by machine learning, and learning methods such as error backpropagation, feedback alignment, direct feedback alignment, synthetic gradient, target prop, difference target prop, and bootstrap can be used.

[0076] When the data 125 on the presence or absence of stroke is a binary value of high or low risk of in-hospital onset, the weighting variables of the learning model can be updated to generate a calculation model M1 so that the information on the risk of in-hospital onset calculated (estimated) based on the input medical record information data matches the output label, where high is 1 and low is 0.

[0077] Fig. 9 is a schematic diagram showing an example of a calculation model M1. The calculation model M1 shown in Fig. 9 is a neural network model having an input layer M1-1, a hidden layer M1-2, and an output layer M1-3. Medical record information data of the training data TD is input to each of the so-called "neurons" of the input layer M1-1. The number of neurons in the hidden layer M1-2 may be more or less than the number of neurons in the input layer M1-1.

[0078] Fig. 10 is a schematic diagram showing another example of the calculation model M1. The calculation model M1 shown in Fig. 10 is a deep neural network model including a convolutional neural network having an input layer M1-1, a convolutional layer M1-2, a pooling layer M1-3, and an output layer M1-4. The convolutional layer M1-2 and the pooling layer M1-3 may each have two or more layers. Medical record information data of the training data TD is input to each neuron of the input layer M1-1.

[0079] The learning processing unit 153 generates or updates a calculation model in which the weights of each neuron in the neural network are learned by performing known machine learning using the training data TD. The learning processing unit 153 may also generate or update a calculation model in which the weights of each neuron in the multi-layered neural network are learned by performing known deep learning using the training data TD.

[0080] 11 shows an example configuration diagram of the present system 100 when predicting information related to a patient's risk of in-hospital onset using machine learning. A predictor made up of a conversion processing unit 152, a risk information calculation unit 102, and a risk information output unit 103 preprocesses the medical record information data as desired, and calculates information related to the patient's risk of in-hospital onset using a trained calculation model M1 with the preprocessed medical record information data as input data.

[0081] The risk information calculation unit 102 executes calculation processing. The risk information calculation unit 102 calculates information regarding the patient's risk of in-hospital onset using the trained calculation model M1 with the acquired medical record information data as input data.

[0082] The calculation process can be executed as desired by the conversion processing unit 152, the risk information output unit 103, and the risk information output unit 103. In the calculation process, when data of a patient's medical record information is acquired from the medical record server 30, information on the patient's risk of in-hospital onset corresponding to the acquired data of the medical record information is calculated.

[0083] (Hospital-onset cerebral infarction algorithm) Figure 5 shows a flowchart of the in-hospital stroke algorithm. When a patient in the hospital shows symptoms that suggest a stroke, medical professionals check the patient's vital signs. Vital signs include items that are traditionally measured on patients, such as blood pressure, pulse rate, and temperature.

[0084] The measured vital signs can be acquired by the medical record information acquisition unit 101. The risk information calculation unit 102 can determine whether the vital signs acquired by the medical record information acquisition unit 101 are normal or abnormal. If the vital signs are determined to be abnormal, the processing unit of this device can notify an in-hospital Rapid Response System (RRS) that can be connected to this device via a network or the like, or can output information that should be notified to the RRS. The RRS is a medical safety system that recognizes precursors in a timely manner and responds quickly to reduce serious adverse events such as unexpected in-hospital cardiac arrest and in-hospital death.

[0085] The processing unit 15 can determine whether or not there is an abnormality in the vital signs based on reference values ​​pre-registered in the memory unit 12, for example, by determining whether at least one of the following applies: blood pressure (BP) less than 80 mmHg, pulse rate (HR) less than 40 beats / minute, and transcutaneous oxygen saturation (SpO2) less than 90%.

[0086] If it is determined that there are no abnormalities in the vital signs, the processing unit 15 can output a request to check the Maria Prehospital Stroke Scale (MPSS). The MPSS is a simple scale that adds stroke severity assessment to the Cincinnati Prehospital Stroke Scale (CPSS), and has the advantage of reducing the time burden on medical professionals such as emergency personnel when conducting on-site evaluations. The user of this device can check the MPSS based on the output confirmation request.

[0087] The MPSS evaluates whether there are any abnormalities in the patient's face, arms, or speech, and scores them out of 5. For the face, symmetry is given a score of 0, and asymmetry is given a score of 1. For the arms, movement of both hands is given a score of 0, one hand moving and pronating is given a score of 1, and one hand dropping or not lifting is given a score of 2. For speech, normal conversation is given a score of 0, unclear speech is given a score of 1, and no speech is given a score of 2. The user of this device can have the medical record information acquisition unit 101 acquire the MPSS score. The risk information calculation unit 102 can notify a predetermined notification destination or output a message that a predetermined notification destination should be notified, depending on the MPSS score.

[0088] If the MPSS is less than 1 point (0 point), the risk information calculation unit 102 can notify the contact information of the attending physician, such as a mobile terminal, or can output information that should be notified to the attending physician.

[0089] An MPSS score of 1 or higher indicates that some symptom has developed, and the risk information calculation unit 102 can calculate the time elapsed since the last symptom-free time based on the medical record information stored in the memory unit of this device. If the time elapsed since the last symptom-free time is more than 4.5 hours, the risk information calculation unit 102 can notify the specialist's contact information, such as a mobile device, and can also output information that should be notified to the specialist. If the time elapsed since the last symptom-free time is 4.5 hours or less, the risk information calculation unit 102 can notify the attending physician's contact information, such as a mobile device, and can also output information that should be notified to the attending physician. The last symptom-free time is the time it is confirmed that the patient is not developing any of the symptoms evaluated in the MPSS.

[0090] The present disclosure is also directed to a method for predicting in-hospital stroke, including obtaining medical record information of a patient at the time of admission or during hospitalization, inputting the medical record information to calculate information related to the patient's risk of in-hospital stroke, and outputting the calculated information related to the patient's risk of in-hospital stroke.

[0091] The present disclosure also relates to an in-hospital stroke prediction program, which includes a medical record information acquisition function that acquires medical record information of a patient at the time of admission to or during hospitalization, a risk information calculation function that calculates information regarding the patient's in-hospital stroke risk when the medical record information is input, and a risk information output function that outputs the calculated information regarding the in-hospital stroke risk. The description of the components included in the in-hospital stroke prediction device described above can be applied to the components included in the in-hospital stroke prediction method and in-hospital stroke prediction program.

[0092] (Embodiment 1: Onset prediction (screening) based on medical record information at the time of hospitalization) The medical records at the time of admission were analyzed to screen the patients for their risk of in-hospital cerebral infarction.

[0093] The medical record information at the time of admission for 54 cases of the in-hospital cerebral infarction group and 270 cases of the non-infarction group was input and acquired by the medical record information acquisition unit 101 .

[0094] The medical record information entered included age, sex, primary disease, medical history, history of cerebral infarction, oral medications, alcohol consumption history, smoking history, family history, test history, and blood test results at the time of admission. Blood test results included white blood cell count, red blood cell count, hematocrit, hemoglobin, platelet count, total protein, albumin, AST, ALT, γGTP, creatinine, urea nitrogen, sodium, potassium, chloride, CRP, uric acid, PT, APTT, D-dimer, BNP, and NT-proBNP. The above medical record information and blood test items are commonly measured when a patient is admitted to the hospital.

[0095] Figure 4 shows a graph showing the incidence rate of in-hospital cerebral infarction by age (predicted risk of in-hospital cerebral infarction) obtained from medical record information at the time of hospitalization. The horizontal axis represents age, and the vertical axis represents the incidence rate of in-hospital cerebral infarction (predicted risk of in-hospital cerebral infarction). Each graph in Figure 4 was created by estimating the value of each item through logistic regression analysis based on the primary disease at the time of admission, age, sex, presence or absence of each medical history (diabetes, hypertension, dyslipidemia, stroke, atrial fibrillation), smoking status, presence or absence of abnormally high or low values ​​for each test value of BNP, PT, APTT, creatinine, BUN, white blood cell count, red blood cell count, hematocrit, hemoglobin, and platelet count, and presence or absence of antithrombotic therapy. Regarding the presence or absence of a patient's history of cerebral infarction, abnormally low BUN, abnormally high BUN, abnormally low WBC, abnormally high WBC, abnormally low PLT, and abnormally high PLT, and the items of age, the sum of the presence or absence of each item represented by two bits, 0 or 1, and the age multiplied by the estimated value of each item was used as x to calculate the formula (1): Probability of in-hospital stroke (in-hospital stroke risk) (%) = e (x) / (1+e (x) ) (1) It was calculated by:

[0096] The intercept β0 was estimated to be -919168, and the estimated values ​​for each item were 0.8612 for history of cerebral infarction, 0.6051 for abnormally low BUN, 0.8164 for abnormally high BUN, -0.0093 for abnormally low WBC, 0.9988 for abnormally high WBC, 0.8166 for abnormally low PLT, 0.4352 for abnormally high PLT, and 0.0281 for age.

[0097] The legends (000000, 000001, etc.) representing each curve on the graph represent a history of cerebral infarction, abnormally low BUN, abnormally high BUN, abnormally low WBC, abnormally high WBC, abnormally low PLT, and abnormally high PLT using two bits, 0 or 1, where 0 means absent and 1 means present. For example, 000000 indicates absence of a history of cerebral infarction, abnormally low BUN, abnormally high BUN, abnormally low WBC, abnormally high WBC, abnormally low PLT, and abnormally high PLT. 000001 indicates absence of a history of cerebral infarction, abnormally low BUN, abnormally high BUN, abnormally low WBC, abnormally high WBC, and abnormally low PLT, but presence of abnormally high PLT. 1010110 indicates that there are no abnormally low BUN values, no abnormally low WBC values, and no abnormally high PLT values, but there is a history of cerebral infarction, and there are abnormally high BUN values, abnormally high WBC values, and abnormally low PLT values.

[0098] Figure 4 shows that the incidence rate of in-hospital cerebral infarction (predicted risk of in-hospital cerebral infarction) increases with age. The curves in legend 1010110 in Figure 4 show that if all conditions are met - a history of cerebral infarction, abnormally high white blood cell and urea nitrogen levels, and abnormally low platelet count - the incidence rate of in-hospital cerebral infarction (predicted risk of in-hospital cerebral infarction) for patients aged 65 or older is 1% or higher. A 1% incidence rate of in-hospital cerebral infarction is 100 times the general incidence rate of 0.01%, making it possible to predict (screen) patients at risk of in-hospital cerebral infarction with high accuracy. In this way, this device can calculate the risk of in-hospital cerebral infarction for each patient.

[0099] (Embodiment 2: Onset prediction based on medical record information during hospitalization) For both the 54 in-hospital stroke group and the 268 non-infarction group, text data was extracted from electronic medical and nursing records from one year prior to the onset date for the in-hospital stroke group and one year prior to discharge date for the non-infarction group. The extracted text data was analyzed using Text Mining Studio® (NTT Data Mathematical Systems). Keywords that may be risk factors for in-hospital stroke, including "atrial fibrillation," "AF," "cancer," "open heart surgery," "catheterization," "endovascular surgery," "heart failure," "cardiac bypass surgery," and "CABG," as well as synonyms, were searched for in both groups, and the number of patients was recorded. Statistical analysis was performed using Fisher's exact test.

[0100] In the analysis of both groups, the incidence and non-incidence groups had 36 / 54 vs. 35 / 268 patients with the words "atrial fibrillation" or "AF," "cancer," and "heart failure" and their synonyms in their medical or nursing records over one year (p=0.007), 18 / 54 vs. 27 / 268 (p<0.0001), and 30 / 54 vs. 106 / 268 (p=0.035). The p-value is the probability of a value more extreme than the observed value when the null hypothesis is assumed to be correct. Text mining analysis revealed that the incidence group had significantly more words containing "atrial fibrillation" or "AF" and "heart failure," which could be a risk factor for in-hospital stroke, and that the medical records of the in-hospital stroke group contained more of these keywords and synonyms. This allows for the evaluation of in-hospital stroke risk through text mining analysis. By combining onset prediction based on medical record information during hospitalization with onset prediction (screening) based on medical record information at the time of admission, it is possible to predict in-hospital onset cerebral infarction with higher accuracy. [Explanation of symbols]

[0101] 100 In-hospital stroke prediction system 10. In-hospital stroke prediction device 20 Network 30 Medical Record Server 101 Medical record information acquisition unit 102 Risk Information Calculation Unit 103 Risk information output unit 12 Storage section 121 Medical Record Information Data 125 Stroke occurrence data 13 Display section 14 Control section 15 Processing section 151 Receiving processing unit 152 Conversion processing section 153 Learning processing unit M1 calculation model TD teacher data

Claims

1. a medical record information acquisition unit that acquires medical record information of a patient at the time of admission to the hospital, during admission, or both; a risk information calculation unit that calculates information regarding the patient's risk of developing an in-hospital stroke when the medical record information is input; and a risk information output unit that outputs information related to the calculated in-hospital stroke risk An in-hospital stroke prediction device comprising:

2. The prediction device according to claim 1 , wherein the medical record information at the time of hospitalization includes age, sex, underlying disease, medical history, whether or not a patient has had a stroke in the past, oral medications, drinking history, smoking history, family history, examination history, blood test results, referral letter, or a combination thereof.

3. The prediction device of claim 1 , wherein the medical record information during hospitalization includes medical records, nursing records, discharge summaries, test results, test findings, surgery records, prescriptions, requests for medical treatment from other departments during hospitalization, or a combination thereof.

4. The prediction device according to claim 1 , wherein the medical record information is medical record information at the time of admission and during the hospitalization.

5. Calculating information about the in-hospital stroke risk includes: Analyzing the text information of the medical record information by text mining to extract words and phrases; and Calculating information regarding the risk of in-hospital stroke using the extracted words and phrases. The prediction device of claim 1 , comprising:

6. Calculating information about the in-hospital stroke risk includes: Analyzing the text information of the medical record information by text mining to extract words and phrases; weighting the extracted phrases; and Using the weighted phrases to calculate information about the risk of in-hospital stroke. The prediction device of claim 1 , comprising:

7. Calculating information about the in-hospital stroke risk includes: Analyzing character data included in the medical record information; and Calculating information regarding the risk of in-hospital stroke using the analyzed character data. The prediction device of claim 1 , comprising:

8. Calculating information about the in-hospital stroke risk includes: Analyzing the text information of the medical record information by text mining to extract words and phrases; Analyzing character data included in the medical record information; and Calculating information regarding the in-hospital stroke risk using the analyzed character data and the extracted words and phrases. The prediction device of claim 1 , comprising:

9. 2. The prediction device of claim 1, wherein calculating the information regarding the in-hospital stroke risk includes using a trained calculation model that has been subjected to machine learning using training data that includes, as input, the subject's medical record information and, as output, information regarding whether the subject has developed a stroke, so that the information regarding the in-hospital stroke risk is calculated when the medical record information is input.

10. The prediction device according to any one of claims 1 to 9; a medical record server in which the medical record information is stored; An in-hospital stroke prediction system comprising:

11. Obtaining medical record information of patients at the time of admission or during their stay at the hospital; inputting the medical record information to calculate information regarding the patient's risk of developing an in-hospital stroke; and outputting information relating to the calculated in-hospital stroke risk; A method for predicting in-hospital stroke, comprising:

12. A medical record information acquisition function that acquires medical record information of patients when they are admitted to or currently hospitalized at a hospital; a risk information calculation function that calculates information about the patient's risk of developing an in-hospital stroke when the medical record information is input; and A risk information output function that outputs information related to the calculated in-hospital stroke risk. In-hospital stroke prediction program, including

Images