CRRT non-planed off-machine early warning and decision support information system

Abstract

The invention discloses a CRRT non-planed offline early warning and decision support information system, and relates to the technical field of data mining. The system comprises a medical record file database, a data analysis server and medical care workstation equipment, and the medical record file database is used for recording medical record files of all historical CRRT online patients. The data analysis server is used for obtaining a prediction model for judging whether a CRRT non-planed off-machine event occurs or not based on successful and failed on-machine medical record files and artificial intelligence algorithm calibration verification modeling; patient attribute information, from the medical workstation equipment, of a to-be-online patient, online scheme attribute information of an online scheme formulated for the to-be-online patient and nursing personnel attribute information of nursing personnel configured for the to-be-online patient are imported into a prediction model, and a prediction result is obtained and fed back to the workstation equipment; therefore, the CRRT non-planned off-machine risk can be accurately identified as early as possible, and the reference information amount is enriched.

Description

Technical Field

[0001] The present invention belongs to the technical field of data mining, and particularly relates to a CRRT unscheduled machine-off warning and decision support information system. Background Art

[0002] Continuous Renal Replacement Therapy (CRRT) refers to a blood purification technology that is carried out continuously for 24 hours or nearly 24 hours every day. Through diffusion and / or convection, solute exchange is carried out continuously and slowly, maximizing the simulation of the solute and water clearance mode of the renal glomerulus and the reabsorption function of the renal tubule, removing excessive water, large and medium molecules, small molecule uremic toxins and metabolic wastes in the blood, while supplementing an equal amount of replacement fluid to maintain the stability of cardiovascular function, so as to replace the damaged renal function and create favorable conditions for the recovery of the body function. CRRT has the advantages of maintaining good hemodynamic stability, maintaining the stability of cerebral perfusion pressure, high solute clearance rate, and preventing the occurrence of disequilibrium syndrome. With the increasing maturity of CRRT technology, its clinical application scope is also expanding, especially playing an important role in the treatment of critically ill diseases such as severe infection, trauma, poisoning and multiple organ failure.

[0003] At present, the machine-on time is one of the important indicators for evaluating the treatment effect of CRRT in patients. However, the unscheduled machine-off caused by the forced interruption of CRRT treatment due to various factors has become a common problem existing in clinical practice. Among them, the unscheduled machine-off refers to the non-artificial planned machine-off without achieving the treatment goal or treatment time. The main reference index for the treatment goal is the treatment dose, that is, the sum of the dehydrated volume and the replacement fluid volume. Research shows that the incidence of unscheduled machine-off in clinical practice is as high as 55% - 66.6%. When critically ill patients receive CRRT treatment, the occurrence of unscheduled machine-off not only directly affects the treatment effect, increases the economic and psychological burden of patients and their families, prolongs the disease recovery period of patients, and even affects the survival rate of patients; at the same time, it also indirectly increases the workload of medical staff and causes waste of medical resources. Therefore, how to accurately identify the risk of CRRT unscheduled machine-off as early as possible, so as to enrich the reference information, assist clinicians to accurately make CRRT machine-on decisions, and then effectively reduce the incidence of CRRT unscheduled machine-off and ensure the CRRT treatment effect is an urgent research topic for those skilled in the art. Summary of the Invention

[0004] The purpose of the present invention is to provide a CRRT unscheduled machine-off warning and decision support information system to solve the problem that the prior art cannot realize early identification of the risk of CRRT unscheduled machine-off and timely warning to effectively assist clinicians in accurately making CRRT machine-on decisions.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] The present invention provides a CRRT unscheduled termination warning and decision support information system, including a medical record database, a data analysis server, and a medical staff workstation device. Among them, the data analysis server is communicatively connected to the medical record database and the medical staff workstation device respectively;

[0007] The medical record database is used to record the medical records of all historical CRRT patients during the machine-on period. Among them, the medical record includes patient attribute information, machine-on plan attribute information, nursing staff attribute information, and the reason for machine termination. The reason for machine termination is divided into scheduled termination and unscheduled termination;

[0008] The data analysis server is used to randomly select an equal number of successfully machine-on medical records and failed machine-on medical records from the medical record database, then sort out negative sample data according to the successfully machine-on medical records, and sort out positive sample data according to the failed machine-on medical records. Finally, all positive sample data and all negative sample data are imported into a machine learning model based on an artificial intelligence algorithm for calibration verification modeling to obtain a prediction model for judging whether a CRRT unscheduled termination event occurs. Among them, a successfully machine-on medical record refers to a medical record whose reason for machine termination is scheduled termination, and a failed machine-on medical record refers to a medical record whose reason for machine termination is unscheduled termination;

[0009] The medical staff workstation device is used to input the patient attribute information of the patient to be machine-on, the machine-on plan attribute information of the machine-on plan formulated for the patient to be machine-on, and the nursing staff attribute information of the nursing staff configured for the patient to be machine-on, and transmit the input data to the data analysis server;

[0010] The data analysis server is also used to import the input data into the prediction model, output the prediction result and feedback it to the medical staff workstation device;

[0011] The medical staff workstation device is also used to determine the CRRT unscheduled termination warning level according to the prediction result and output and display it, so as to provide reference auxiliary information for medical staff when making a decision on whether to perform CRRT machine-on for the patient to be machine-on.

[0012] Based on the above invention content, a new solution for risk early warning of unplanned termination of CRRT based on artificial intelligence algorithm to achieve decision support is provided, including a medical record database, a data analysis server and a medical staff workstation device. Among them, the medical record database is used to record the medical records of all historical CRRT patients during the machine-on period. The data analysis server is used to calibrate and verify the modeling based on the successful and failed machine-on medical records and the artificial intelligence algorithm to obtain a prediction model for judging whether an unplanned termination event of CRRT occurs, and import the patient attribute information of the patient to be machine-on from the medical staff workstation device, the machine-on plan attribute information of the machine-on plan formulated for this patient to be machine-on, and the nurse attribute information of the nurse configured for this patient to be machine-on into the prediction model, obtain the prediction result and feedback it to the medical staff workstation device, so that it can determine the early warning level of unplanned termination of CRRT according to the prediction result and output and display it. In this way, the risk of unplanned termination of CRRT can be accurately identified as early as possible, the reference information amount can be enriched, the clinical doctor can be assisted to accurately make the CRRT machine-on decision, and then the incidence rate of unplanned termination of CRRT can be effectively reduced, the CRRT treatment effect can be guaranteed, and it is convenient for practical application and popularization.

[0013] In a possible design, the patient attribute information includes the label values of the patient's age, gender, patient disease-related parameters and / or patient coagulation-related parameters. Among them, the patient disease-related parameters include the label values of the machine-on position, diastolic blood pressure, systolic blood pressure and / or APACHE II score. The machine-on position is divided into left lateral position, supine position and right lateral position. The patient coagulation-related parameters include prothrombin time, activated partial thromboplastin time, fibrinogen and / or platelet count.

[0014] In a possible design, the machine-on plan attribute information includes the planned machine-on duration, machine-on anticoagulation-related parameters, machine-on blood purification-related parameters and / or machine-on catheter-related parameters. Among them, the machine-on anticoagulation-related parameters include the label values of the number of blood products infused and / or the types of anticoagulants. The machine-on blood purification-related parameters include the label values of the treatment mode, filter type, pre-dilution fluid volume, post-dilution fluid volume, blood flow rate, arterial pressure and / or transmembrane pressure. The machine-on catheter-related parameters include the label values of the catheterization site, catheterization duration and / or catheter specification.

[0015] In a possible design, the nurse attribute information includes the working years of the nurse, the label value of the nurse's gender and / or the label value of the nurse's position level.

[0016] In a possible design, negative sample data is sorted out according to the successful machine-on medical records, and positive sample data is sorted out according to the failed machine-on medical records, including:

[0017] Randomly sort all the selected medical record files to obtain a sequence of medical record files;

[0018] For each medical record file in the medical record file sequence, if the corresponding logout reason is planned logout, determine that the occurrence label value of the corresponding CRRT unplanned logout event is zero; otherwise, determine that the occurrence label value of the corresponding CRRT unplanned logout event is one;

[0019] For each parameter in the medical record file except the logout reason, calculate the correlation coefficient between the corresponding parameter and the CRRT unplanned logout event according to the occurrence label value sequence and the corresponding parameter value sequence that are in one-to-one correspondence with the medical record file sequence;

[0020] Arrange all the parameters in the medical record file except the logout reason in descending order of the correlation coefficient to obtain a parameter sequence;

[0021] Select the first N parameters from the parameter sequence as key parameters respectively, where N represents a positive integer greater than or equal to 5 and less than or equal to 15;

[0022] According to the successfully logged-in medical record files, organize and obtain negative sample data in the following way: Use the parameter values of all the key parameters in the successfully logged-in medical record files as the first model input item, and use the occurrence label value of the CRRT unplanned logout event corresponding to the successfully logged-in medical record files as the first model output item, and then use this first model input item and this first model output item together as negative sample data;

[0023] According to the failed logged-in medical record files, organize and obtain positive sample data in the following way: Use the parameter values of all the key parameters in the failed logged-in medical record files as the second model input item, and use the occurrence label value of the CRRT unplanned logout event corresponding to the failed logged-in medical record files as the second model output item, and then use this second model input item and this second model output item together as positive sample data.

[0024] In a possible design, for each parameter in the medical record file except the logout reason, calculate the correlation coefficient between the corresponding parameter and the CRRT unplanned logout event according to the occurrence label value sequence and the corresponding parameter value sequence that are in one-to-one correspondence with the medical record file sequence, including:

[0025] For any parameter in the medical record file except the logout reason, according to all the selected medical record files, extract the corresponding parameter value sequence S1 that is in one-to-one correspondence with the medical record file sequence, and perform a normal distribution KS test on the corresponding parameter value sequence S1 to calculate the first test statistic p value p1;

[0026] Perform a normal distribution KS test on the sequence of occurrence tag values S2 that corresponds one-to-one with the medical record file sequence to calculate the second test statistic p-value p2;

[0027] If both the first test statistic p-value p1 and the second test statistic p-value p2 are greater than the preset threshold, then calculate the correlation coefficient r between any of the parameters and the CRRT unplanned machine-off event according to the following formula q :

[0028]

[0029] In the formula, i represents a positive integer, x i represents the i-th parameter value in S1, y i represents the i-th occurrence tag value in S2, represents the mean value of S1, represents the mean value of S2.

[0030] In a possible design, after selecting the first N parameters from the parameter sequence as the key parameters respectively, import the input data into the prediction model, including:

[0031] Extract the parameter values of all key parameters from the input data, and import the extraction result into the prediction model.

[0032] [[ID=2८]]In a possible design, the data analysis server is further configured to apply the prediction model and combine with the optimization algorithm to optimize the machine-on plan formulated for the patient to be machine-on according to the patient attribute information of the patient to be machine-on in the input data and the caregiver attribute information of the caregiver configured for the patient to be machine-on, obtain the machine-on plan and the optimal search result for minimizing the CRRT unplanned machine-off incidence rate, and then when the second CRRT unplanned machine-off incidence rate corresponding to the optimal search result is lower than the first CRRT unplanned machine-off incidence rate obtained according to the prediction result, use the optimal search result as the recommended machine-on plan formulated for the patient to be machine-on, and finally transmit the recommended machine-on plan and the second CRRT unplanned machine-off incidence rate to the medical staff workstation device together;

[0033] The medical staff workstation device is further configured to output and display the recommended machine-on plan, and determine and output and display a new CRRT unplanned machine-off warning level according to the second CRRT unplanned machine-off incidence rate, so as to provide reference auxiliary information for medical staff when making a decision on whether to perform CRRT machine-on for the patient to be machine-on.

[0034] In a possible design, according to the patient attribute information of the patient to be connected to the machine in the input data and the caregiver attribute information of the caregiver configured for the patient to be connected to the machine, a prediction model is applied and combined with an optimization algorithm to optimize the connection plan formulated for the patient to be connected to the machine, and an optimized search result for minimizing the incidence of unplanned disconnection of CRRT is obtained, including the following steps S301 to S309:

[0035] S301. Initialize the optimization algorithm parameters including the maximum number of iterations T, and randomly generate an initial search value array of the set of parameters to be optimized, and then execute step S302. Among them, the set of parameters to be optimized includes all parameters in the connection plan attribute information of the connection plan formulated for the patient to be connected to the machine. The initial search value array y of the set of parameters to be optimized 0 is expressed as follows:

[0036]

[0037] In the formula, d″ represents a positive integer less than or equal to D″, and D″ represents the total number of parameters in the set of parameters to be optimized, represents the initial search value corresponding to the d″-th parameter in the set of parameters to be optimized, u c,d″ represents the upper limit of the parameter search space corresponding to the d″-th parameter, l c,d″ represents the lower limit of the parameter search space corresponding to the d″-th parameter, and rand(0, 1) represents a pure decimal random generation function;

[0038] S302. Import the patient attribute information of the patient to be connected to the machine in the input data and the caregiver attribute information of the caregiver configured for the patient to be connected to the machine together with the initial search value array of the set of parameters to be optimized into the prediction model, output the first confidence level for determining the occurrence of an unplanned disconnection event of CRRT, and use this first confidence level as the fitness corresponding to the initial search value array, and then execute step S303;

[0039] S303. Take the initial search value array of the set of parameters to be optimized as the current optimal search value array, initialize and set the current iteration number t′ = 0, and also initialize and set the forbidden change countdown value of each parameter in the set of parameters to be optimized to zero, and then execute step S304;

[0040] S304. Based on the current optimal search value array, perform independent increase processing and decrease processing on each parameter in the set of parameters to be optimized and with a current forbidden change countdown value of zero, to obtain 2×D″′ new search value arrays of the set of parameters to be optimized, and then execute step S305. Among them, D″′ represents the total number of parameters in the set of parameters to be optimized and with a current forbidden change countdown value of zero, and 1 ≤ D″′ ≤ D″;

[0041] S305. For each of the 2×D″′ new search value arrays, import the patient attribute information of the patient to be connected to the machine in the input data and the caregiver attribute information of the caregiver configured for the patient to be connected to the machine together with the corresponding array into the prediction model, output the second confidence level for determining the occurrence of the CRRT unscheduled disconnection event, and use this second confidence level as the fitness of the corresponding array, and then execute step S306;

[0042] S306. For each array, subtract the fitness of the corresponding array from the fitness of the array corresponding to the current optimal search value array to obtain the fitness difference value of the corresponding array. When the fitness difference value is greater than zero and is the largest fitness difference value in this time, update the forbidden change countdown value of the corresponding unique change parameter to a positive integer value that is positively correlated with this fitness difference value, and then execute step S307, where the unique change parameter refers to the parameter in the set of parameters to be optimized and used to obtain the corresponding array through an increase process or a decrease process;

[0043] S307. Determine whether there is any parameter in the set of parameters to be optimized whose current forbidden change countdown value is zero. If not, decrement the current forbidden change countdown value of each parameter in the set of parameters to be optimized by 1, and then return to execute step S307. Otherwise, execute step S308;

[0044] S308. Determine whether there is any fitness difference value greater than zero among the fitness difference values of each array. If so, update the current optimal search value array to the array in the 2×D″′ new search value arrays that corresponds to the minimum fitness, and then execute step S309. Otherwise, directly execute step S309;

[0045] S309. Increment the current iteration count t′ by 1, and determine whether the current iteration count t′ has reached the maximum iteration count T. If so, use the current optimal search value array as the optimal search result for formulating the connection plan for the patient to be connected to the machine and minimizing the CRRT unscheduled disconnection rate. Otherwise, return to execute step S304.

[0046] In a possible design, determining the CRRT unscheduled disconnection warning level according to the prediction result includes:

[0047] Extract the confidence level for the input data and determining the occurrence of the CRRT unscheduled disconnection event according to the prediction result, and use this confidence level as the first CRRT unscheduled disconnection rate;

[0048] According to all preset probability ranges corresponding one by one to all early warning levels of unplanned discontinuation of CRRT, if it is found that the incidence rate of the first unplanned discontinuation of CRRT is within the preset probability range corresponding to a certain early warning level of unplanned discontinuation of CRRT, then take this certain early warning level of unplanned discontinuation of CRRT as the early warning level of unplanned discontinuation of CRRT determined according to the prediction result.

[0049] Beneficial effects of the above solution:

[0050] (1) The present invention provides a new solution for risk early warning of unplanned discontinuation of CRRT based on artificial intelligence algorithm to achieve decision support, including a medical record database, a data analysis server and a medical staff workstation device. Among them, the medical record database is used to record the medical records of all historical CRRT patients, and the data analysis server is used to calibrate and verify the modeling based on the successful and failed medical records of CRRT and the artificial intelligence algorithm to obtain a prediction model for judging whether an unplanned discontinuation of CRRT event occurs, and import the patient attribute information of the patient to be connected to the machine from the medical staff workstation device, the attribute information of the connection plan for the patient to be connected to the machine, and the attribute information of the nursing staff configured for the patient to be connected to the machine into the prediction model, obtain the prediction result and feedback it to the medical staff workstation device, so that it can determine the early warning level of unplanned discontinuation of CRRT according to the prediction result and output and display it. In this way, the risk of unplanned discontinuation of CRRT can be accurately identified as early as possible, the reference information amount can be enriched, the clinical doctor can be assisted to accurately make a decision on CRRT connection, and further effectively promote the reduction of the incidence rate of unplanned discontinuation of CRRT, ensure the treatment effect of CRRT, and facilitate practical application and promotion;

[0051] (2) The prediction model and the optimization algorithm can also be combined to optimize the best patient connection plan and the nursing staff allocation plan, further enriching the reference information amount;

[0052] (3) A new heuristic algorithm step is also provided to optimize the connection plan. During the optimization process, different iterative lockings of the search results in each best neighborhood search direction can be performed based on different fitness difference values, which is conducive to quickly searching for the optimal connection plan attribute information. By first quantitatively increasing / decreasing the adjustable parameters and then randomly increasing / decreasing them in an indefinite amount, it is also possible to avoid falling into local optimal solutions, which is further conducive to quickly and accurately obtaining the optimal search result. Description of the Drawings

[0053] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0054] Figure 1 FIG. is a schematic structural diagram of the CRRT unscheduled machine stop warning and decision support information system provided by the embodiment of the present invention. Detailed implementation manners

[0055] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the present invention in combination with the accompanying drawings and the description of the embodiments or the prior art. Obviously, the following description of the structure of the accompanying drawings is only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other embodiments can be obtained based on these embodiments. It should be noted here that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation to the present invention.

[0056] It should be understood that although terms such as first and second etc. may be used herein to describe various objects, these objects should not be limited by these terms. These terms are only used to distinguish one object from another. For example, the first object can be called the second object, and similarly, the second object can be called the first object, without departing from the scope of the exemplary embodiments of the present invention.

[0057] It should be understood that for the term "and / or" that may appear in this document, it is only a description of the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, or A and B exist simultaneously, etc. Three situations; again, for example, A, B and / or C can mean the existence of any one of A, B and C or any combination of them; for the term " / and" that may appear in this document, it is a description of another association object relationship, indicating that two relationships can exist. For example, A / and B can mean: A exists alone or A and B exist simultaneously, etc. Two situations; in addition, for the character " / " that may appear in this document, generally it means that the front and rear associated objects are an "or" relationship.

[0058] Embodiment

[0059] Such as Figure 1As shown, the CRRT unscheduled machine shutdown warning and decision support information system provided in this embodiment includes, but is not limited to, a medical record database, a data analysis server, a medical staff workstation device, etc. Among them, the data analysis server is communicatively connected to the medical record database and the medical staff workstation device respectively.

[0060] The medical record database is used to record the medical records of all historical CRRT patients during the treatment. Among them, the medical records include, but are not limited to, patient attribute information, treatment plan attribute information, nurse attribute information, and reasons for terminating the treatment, etc. The reasons for terminating the treatment are divided into planned termination and unplanned termination. The aforementioned historical CRRT patients during the treatment refer to the patients who have undergone CRRT treatment historically. Therefore, the corresponding medical records can be routinely recorded and stored in the medical record database. Specifically, the patient attribute information includes, but is not limited to, patient age, label values of patient gender (for example, using the value "1" to represent male and the value "0" to represent female), patient disease-related parameters, and / or patient coagulation-related parameters, etc. Among them, the patient disease-related parameters include, but are not limited to, label values of the treatment position, diastolic blood pressure, systolic blood pressure, and / or APACHE II score (the full English name of APACHE is Acute Physiology and Chronic Health Evaluation, which is translated into Chinese as Acute Physiological and Chronic Health Score), etc. The treatment positions are divided into left lateral position (the corresponding label value can be exemplified as the value "-1"), supine position (the corresponding label value can be exemplified as the value "0"), and right lateral position (the corresponding label value can be exemplified as the value "1"), etc. The patient coagulation-related parameters include, but are not limited to, prothrombin time, activated partial thromboplastin time, fibrinogen, and / or platelet count, etc. The treatment plan attribute information includes, but is not limited to, planned treatment duration, anticoagulation-related parameters during the treatment, blood purification-related parameters during the treatment, and / or catheter-related parameters during the treatment, etc. Among them, the anticoagulation-related parameters during the treatment include, but are not limited to, the number of types of blood products transfused and / or label values of anticoagulant types, etc. The blood purification-related parameters during the treatment include, but are not limited to, label values of treatment modes, label values of filter types, pre-dilution fluid volume, post-dilution fluid volume, blood flow rate, arterial pressure, and / or transmembrane pressure, etc. The catheter-related parameters during the treatment include, but are not limited to, label values of catheter insertion sites, catheter insertion duration, and / or label values of catheter specifications, etc. The nurse attribute information includes, but is not limited to, the working years of the nurse, label values of nurse gender (for example, using the value "1" to represent male and the value "0" to represent female), and / or label values of the position levels where the nurse is located, etc.The aforementioned types of anticoagulants can specifically but not limited to be divided into heparin anticoagulants, low molecular weight heparin anticoagulants, and non - use of anticoagulants, etc.; the aforementioned treatment modes can specifically but not limited to be divided into continuous venovenous hemofiltration mode (CVVH), continuous venovenous hemodialysis mode (CVVHD), continuous venovenous hemodiafiltration mode (CVVHDF), slow continuous ultrafiltration mode (SCUF), continuous plasma filtration adsorption mode (CPFA), endotoxin adsorption mode, plasma exchange mode (PE), and double - membrane plasma exchange mode (DFPP), etc.; the aforementioned filter types can specifically but not limited to be divided into AV600S type filters, AV600S + HA330 type filters, HF1200 type filters, and HF1200 + HA330 type filters, etc.; the aforementioned catheter insertion sites can specifically but not limited to be divided into right femoral vein site, left femoral vein site, and other sites, etc.; the aforementioned catheter specifications can specifically but not limited to be divided into 11.5F×16cm specifications, 11.5F×20cm specifications, 7F×20cm specifications, and triple - lumen 12F×16cm specifications, etc.; the aforementioned job levels of nursing staff can specifically but not limited to be divided into three levels: primary, intermediate, and senior. Each level is further subdivided into different grades. For example, the senior level is further subdivided into N3 level and N4 level. Each specific term in the medical record file is a common term in the medical and nursing field, and its specific meaning will not be elaborated here. In addition, the medical record file database can specifically but not limited to be conventionally implemented using a relational database.

[0061] The data analysis server is used to randomly select an equal number of successfully connected medical record files and failed connected medical record files from the medical record file database. Then, negative sample data is sorted out according to the successfully connected medical record files, and positive sample data is sorted out according to the failed connected medical record files. Finally, all positive sample data and all negative sample data are imported into a machine learning model based on an artificial intelligence algorithm for calibration verification modeling to obtain a prediction model for judging whether a CRRT unplanned disconnection event occurs. Among them, the successfully connected medical record file refers to the medical record file whose disconnection reason is planned disconnection, and the failed connected medical record file refers to the medical record file whose disconnection reason is unplanned disconnection. The specific process of randomly selecting successfully connected medical record files and failed connected medical record files can be conventionally implemented in combination with existing pseudo - random algorithms. For example, assume that the total number of successfully connected medical record files in the medical record file database is N su , then a pseudo - random algorithm can be used to generate 100 different numbers within the interval [1, N suRandom positive integers within [], and then select the successful on-machine medical record files with the serial numbers being these random positive integers to obtain 100 successful on-machine medical record files; similarly, 100 failed on-machine medical record files can also be obtained. The purpose of the aforementioned requirement that the number of successful on-machine medical record files and failed on-machine medical record files is equal (for example, both are 100) is to make the distribution of each result relatively uniform in all subsequent sample data, otherwise the trained model cannot achieve the basic prediction function. Considering that there are many model input item parameters available in the medical record files, and different parameters will surely have different impacts on the unplanned machine shutdown, in order to ensure that a highly accurate prediction model can be trained subsequently and to achieve the purpose of minimizing the computing resource requirements through dimensionality reduction, preferably, negative sample data is sorted out according to the successful on-machine medical record files, and positive sample data is sorted out according to the failed on-machine medical record files, including but not limited to the following steps S101 to S107.

[0062] S101. Randomly sort all the selected medical record files to obtain a medical record file sequence.

[0063] In step S101, all the aforementioned selected medical record files include all the randomly selected successful on-machine medical record files and all the failed on-machine medical record files (the two cases are equal in number) from the medical record file database.

[0064] S102. For each medical record file in the medical record file sequence, if the corresponding machine shutdown reason is planned machine shutdown, determine that the occurrence label value of the corresponding CRRT unplanned machine shutdown event is zero, otherwise determine that the occurrence label value of the corresponding CRRT unplanned machine shutdown event is one.

[0065] S103. For each parameter in the medical record file except the machine shutdown reason, calculate the correlation coefficient between the corresponding parameter and the CRRT unplanned machine shutdown event according to the occurrence label value sequence and the corresponding parameter value sequence that are in one-to-one correspondence with the medical record file sequence.

[0066] In step S103, the aforementioned parameters are respectively the label value of the patient's age, the label value of the patient's gender, the patient's disease-related parameters, the patient's coagulation-related parameters, the planned on-machine duration, the on-machine anticoagulation-related parameters, the on-machine blood purification-related parameters, the on-machine catheter-related parameters, the working years of the nursing staff, the label value of the nursing staff's gender, and / or the label value of the nursing staff's position level, etc. Specifically, for each parameter in the medical record file except the machine shutdown reason, calculate the correlation coefficient between the corresponding parameter and the CRRT unplanned machine shutdown event according to the occurrence label value sequence and the corresponding parameter value sequence that are in one-to-one correspondence with the medical record file sequence, including but not limited to the following steps S1031 to S1033.

[0067] S1031. For any parameter in the medical record file other than the reason for machine shutdown, based on all the selected medical record files, extract the corresponding parameter value sequence S1 that corresponds one by one to the medical record file sequence, and perform a Kolmogorov-Smirnov (KS) test for normal distribution on this corresponding parameter value sequence S1 to calculate the first test statistic p-value p1.

[0068] In the step S1031, the normal distribution KS test is the Kolmogorov-Smirnov test, which is an existing statistical test method used to determine whether the observed values of a set of data conform to a specific distribution. Therefore, it can be conventionally modified to obtain the first test statistic p-value p1.

[0069] S1032. Perform the normal distribution KS test on the occurrence label value sequence S2 that corresponds one by one to the medical record file sequence to calculate the second test statistic p-value p2.

[0070] S1033. If both the first test statistic p-value p1 and the second test statistic p-value p2 are greater than the preset threshold, then calculate the correlation coefficient r between the any parameter and the unplanned machine shutdown event of CRRT according to the following formula q :

[0071]

[0072] In the formula, i represents a positive integer, x i represents the i-th parameter value in S1, y i represents the i-th occurrence label value in S2, represents the mean value of S1, represents the mean value of S2.

[0073] In the step S1033, the preset threshold can be exemplified as 0.05.

[0074] S104. Arrange all the parameters in the medical record file other than the reason for machine shutdown in descending order of the correlation coefficient to obtain a parameter sequence.

[0075] S105. Select the first N parameters from the parameter sequence as the key parameters respectively, where N represents a positive integer greater than or equal to 5 and less than or equal to 15.

[0076] In the step S105, N can be exemplified as 10.

[0077] S106. According to the successfully connected medical record files, negative sample data is sorted out in the following manner: taking the parameter values of all key parameters in the successfully connected medical record files as the first model input items, and taking the occurrence label value of the CRRT unplanned disconnection event corresponding to the successfully connected medical record files as the first model output item, and then taking this first model input item and this first model output item together as the negative sample data.

[0078] S107. According to the failed connected medical record files, positive sample data is sorted out in the following manner: taking the parameter values of all key parameters in the failed connected medical record files as the second model input items, and taking the occurrence label value of the CRRT unplanned disconnection event corresponding to the failed connected medical record files as the second model output item, and then taking this second model input item and this second model output item together as the positive sample data.

[0079] The aforementioned artificial intelligence algorithm is a core artificial intelligence algorithm that specifically studies how a computer simulates or implements human learning behaviors to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve its own performance, and is the fundamental way to make a computer intelligent; specifically, the artificial intelligence algorithm can but is not limited to specifically adopting machine learning algorithms based on support vector machines, stochastic gradient descent methods, multivariate linear regression, multi-layer perceptrons, decision trees, backpropagation neural networks, or radial basis function networks, etc., so as to quickly and accurately find the rules in the data. Therefore, based on all the positive sample data and all the negative sample data, through a conventional calibration verification modeling process (specifically including the calibration process and verification process of the model, that is, first comparing the model simulation results with the measured data, and then adjusting the model parameters according to the comparison results to make the simulation results coincide with the actual situation), the prediction model for judging whether a CRRT unplanned disconnection event occurs is trained. In addition, in the calibration verification modeling process of the prediction model, it is preferred to use a Bayesian optimization algorithm based on a tree structure to optimize the model parameters; the data analysis server can specifically adopt an existing server structure.

[0080] The medical staff workstation device is used to input the patient attribute information of the patient to be connected, the machine connection plan attribute information of the machine connection plan formulated for this patient to be connected, and the nursing staff attribute information of the nursing staff configured for this patient to be connected, and transmit the input data to the data analysis server. The aforementioned machine connection plan can be manually formulated by the attending doctor of the patient to be connected; the aforementioned nursing staff can be automatically assigned based on the existing technology or can also be assigned by this attending doctor; the aforementioned input data can be input manually or can also be input automatically. In addition, the medical staff workstation device can specifically be implemented by using an existing software and hardware structure.

[0081] The data analysis server is further configured to import the input data into the prediction model, output a prediction result, and feedback it to the medical workstation device. The foregoing prediction result includes the following two situations: (1) It is determined that a CRRT unscheduled machine stop event will occur for the input data (at this time, a numerical value one is output to represent it) and a confidence level is given; (2) It is determined that a CRRT unscheduled machine stop event will not occur for the input data (at this time, a numerical value zero is output to represent it) and a confidence level is given. In addition, if the first N parameters have been selected from the parameter sequence as key parameters respectively, in order to ensure the consistency of the model input items in the model training stage and the model application stage, preferably, after the first N parameters are selected from the parameter sequence as key parameters respectively, importing the input data into the prediction model includes, but is not limited to: extracting the parameter values of all key parameters from the input data, and importing the extraction result into the prediction model.

[0082] The medical workstation device is further configured to determine the CRRT unscheduled machine stop warning level according to the prediction result and output and display it, so as to provide reference auxiliary information for medical staff when making a decision on whether to perform CRRT machine start for the patient to be machine-started. Since the prediction result will give the confidence level of determining that a CRRT unscheduled machine stop event occurs or the confidence level of determining that a CRRT unscheduled machine stop event does not occur, the difference between the former confidence level or 1 minus the latter confidence level can be used as the estimated CRRT unscheduled machine stop incidence rate for the input data, and then the CRRT unscheduled machine stop warning level can be determined according to this CRRT unscheduled machine stop incidence rate. Specifically, determining the CRRT unscheduled machine stop warning level according to the prediction result includes, but is not limited to, the following steps S201 to S202.

[0083] S201. Extract the confidence level of determining that a CRRT unscheduled machine stop event occurs for the input data according to the prediction result, and use this confidence level as the first CRRT unscheduled machine stop incidence rate.

[0084] S202. According to all preset probability ranges corresponding to all CRRT unscheduled machine stop warning levels one by one, if it is found that the first CRRT unscheduled machine stop incidence rate is within the preset probability range corresponding to a certain CRRT unscheduled machine stop warning level, then use this certain CRRT unscheduled machine stop warning level as the CRRT unscheduled machine stop warning level determined according to the prediction result.

[0085] In the step S202, all the early warning levels of unplanned discontinuation of CRRT specifically include, but are not limited to, risk early warning (the corresponding preset probability range is, for example, less than 5%), low-risk early warning level (the corresponding preset probability range is, for example, 5% - 15%), medium-risk early warning level (the corresponding preset probability range is, for example, 15% - 30%), and high-risk early warning level (the corresponding preset probability range is, for example, more than 30%). For example, if the incidence rate of unplanned discontinuation of the first CRRT is 45%, the high-risk early warning level can be determined as the early warning level of unplanned discontinuation of CRRT determined according to the prediction result and output for display, so as to provide reference auxiliary information for medical staff when making a decision on whether to perform CRRT on the patient to be connected to the machine. Thus, the risk of unplanned discontinuation of CRRT can be accurately identified as early as possible, so as to enrich the reference information volume, assist clinicians in accurately making a decision on CRRT connection to the machine, and further effectively promote the reduction of the incidence rate of unplanned discontinuation of CRRT, ensure the CRRT treatment effect, and facilitate practical application and promotion.

[0086] Preferably, the data analysis server is further configured to optimize the CRRT treatment plan for the to-be-treated patient according to the patient attribute information of the to-be-treated patient in the input data and the caregiver attribute information of the caregiver assigned to the to-be-treated patient, by applying the prediction model and combining with an optimization algorithm, so as to obtain the treatment plan and the optimal search result for minimizing the incidence of unplanned CRRT discontinuation. Then, when the second incidence of unplanned CRRT discontinuation corresponding to the optimal search result is lower than the first incidence of unplanned CRRT discontinuation obtained according to the prediction result, the optimal search result is used as the recommended CRRT treatment plan for the to-be-treated patient. Finally, the recommended CRRT treatment plan and the second incidence of unplanned CRRT discontinuation are transmitted to the medical workstation device together. The medical workstation device is further configured to output and display the recommended CRRT treatment plan, and determine and output a new warning level for unplanned CRRT discontinuation according to the second incidence of unplanned CRRT discontinuation, so as to provide reference and auxiliary information for medical staff when making a decision on whether to perform CRRT treatment for the to-be-treated patient. The aforementioned optimization algorithm can be, but is not limited to, a particle swarm optimization algorithm, a Newton optimization algorithm, a genetic optimization algorithm, a grey wolf optimization algorithm, a whale optimization algorithm, or a tuna school optimization algorithm, etc. In this way, the best patient treatment plan can be optimized, and the reference information can be further enriched. To obtain the optimal search result quickly and accurately, preferably, according to the patient attribute information of the to-be-treated patient in the input data and the caregiver attribute information of the caregiver assigned to the to-be-treated patient, applying the prediction model and combining with an optimization algorithm to optimize the CRRT treatment plan for the to-be-treated patient, obtaining the treatment plan and the optimal search result for minimizing the incidence of unplanned CRRT discontinuation, including the following steps S301 - S309.

[0087] S301. Initialize the parameters of the optimization algorithm including the maximum number of iterations T, and randomly generate an initial search value array of the set of parameters to be optimized. Then, execute step S302, where the set of parameters to be optimized includes all parameters in the treatment plan attribute information of the CRRT treatment plan for the to-be-treated patient, and the initial search value array y 0 is expressed as follows:

[0088]

[0089] In the formula, d″ represents a positive integer less than or equal to D″, and D″ represents the total number of parameters in the set of parameters to be optimized. represents the initial search value corresponding to the d″-th parameter in the set of parameters to be optimized, u c,d″Denote the upper limit of the parameter search space corresponding to the d″-th parameter, l c,d″ Denote the lower limit of the parameter search space corresponding to the d″-th parameter, and rand(0, 1) represents a fractional random generation function.

[0090] In the step S301, the upper limit of the parameter search space and the lower limit of the parameter search space can be obtained through routine statistics of all selected medical record files.

[0091] S302. Import the patient attribute information of the patient to be connected to the machine and the caregiver attribute information of the caregiver configured for the patient to be connected to the machine in the input data, together with the initial search value array of the set of parameters to be optimized, into the prediction model, output the first confidence level for determining the occurrence of an unplanned discontinuation event of CRRT, and use this first confidence level as the fitness corresponding to the initial search value array, and then execute step S303.

[0092] In the step S302, considering that some parameters need to be integers (such as the label value of the treatment mode or the label value of the filter type, etc.), so before importing the initial search value array as one of the model input items into the model, it is necessary to round the corresponding parameter values. For example, round the label value of the treatment mode. Similarly, if the first N parameters have been selected from the parameter sequence as the key parameters respectively, in order to ensure the consistency of the model input items in the model training stage and the model application stage, preferably, after selecting the first N parameters from the parameter sequence as the key parameters respectively, import the patient attribute information of the patient to be connected to the machine and the caregiver attribute information of the caregiver configured for the patient to be connected to the machine in the input data, together with the initial search value array of the set of parameters to be optimized, into the prediction model, including but not limited to: extract the parameter values of all key parameters from the first import data, and import the extraction result into the prediction model, where the first import data includes the patient attribute information of the patient to be connected to the machine and the caregiver attribute information of the caregiver configured for the patient to be connected to the machine in the input data, together with the initial search value array of the set of parameters to be optimized. In addition, the first confidence level is the predicted incidence of unplanned discontinuation of CRRT for the first import data.

[0093] S303. Take the initial search value array of the set of parameters to be optimized as the current optimal search value array, initialize the current iteration count t′ = 0, and also initialize the forbidden change countdown value of each parameter in the set of parameters to be optimized to zero, and then execute step S304.

[0094] S304. Based on the current optimal search value array, perform independent increase processing and decrease processing on each parameter in the set of parameters to be optimized and with the current forbidden change countdown value being zero, to obtain 2×D″′ new search value arrays of the set of parameters to be optimized, and then execute step S305, where D″′ represents the total number of parameters in the set of parameters to be optimized and with the current forbidden change countdown value being zero, and 1≤D″′≤D″.

[0095] In step S304, the increase processing or the decrease processing needs to be carried out within the corresponding parameter search space, and can be a quantitative step - by - step increase / decrease, or an indefinite random increase / decrease. Additionally, it is also possible to first perform quantitative step - by - step increase / decrease on each parameter in the set of parameters to be optimized and with the current forbidden change countdown value being zero, and then if it is found that the update of the current optimal search value array is not completed (i.e., step S309 is directly executed in the subsequent step S308), then perform indefinite random increase / decrease on each parameter in the set of parameters to be optimized and with the current forbidden change countdown value being zero, so as to avoid falling into a local optimal solution. For example, if there are the following six parameters in the set of parameters to be optimized: parameter A, parameter B, parameter C, parameter D, parameter E, and parameter F, where the current forbidden change countdown values of parameter A, parameter B, parameter E, and parameter F are zero respectively (i.e., D″′ takes the value of 4), then based on the current optimal search value array, independent increase processing and decrease processing can be performed on parameter A respectively to obtain two different new search value arrays of the set of parameters to be optimized (one obtained based on the increase processing and the other based on the decrease processing); based on the current optimal search value array, independent increase processing and decrease processing can be performed on parameter B respectively to obtain two different new search value arrays of the set of parameters to be optimized; based on the current optimal search value array, independent increase processing and decrease processing can be performed on parameter E respectively to obtain two different new search value arrays of the set of parameters to be optimized; based on the current optimal search value array, independent increase processing and decrease processing can be performed on parameter F respectively to obtain two different new search value arrays of the set of parameters to be optimized; thus, 2×4 = 8 new search value arrays can be obtained, and each array represents a neighborhood search direction in the search space.

[0096] S305. For each array in the 2×D″′ new search value arrays, import the patient attribute information of the patient to be connected to the machine in the input data and the caregiver attribute information of the caregiver configured for this patient to be connected to the machine together with the corresponding array into the prediction model, output the second confidence level for determining the occurrence of the CRRT unplanned disconnection event, and use this second confidence level as the fitness of the corresponding array, and then execute step S306.

[0097] In the step S305, the specific technical details can be obtained by referring to the conventional derivation in the foregoing step S302, and will not be elaborated here.

[0098] S306. For each of the arrays, subtract the fitness of the corresponding array from the fitness of the array corresponding to the current optimal search value to obtain the fitness difference value of the corresponding array. When the fitness difference value is greater than zero and is the largest fitness difference value in this time, update the freeze countdown value of the corresponding unique variable parameter to a positive integer value positively correlated with the fitness difference value, and then execute step S307, where the unique variable parameter refers to the parameter in the set of parameters to be optimized and used to obtain the corresponding array through increase processing or decrease processing.

[0099] In the step S306, that the fitness difference value is greater than zero and is the largest fitness difference value in this time means that the best search value array has been obtained in the neighborhood search direction corresponding to the corresponding parameter this time. Therefore, it is necessary to update the freeze countdown value of the corresponding parameter to a positive integer value positively correlated with the fitness difference value to temporarily lock the search result in this neighborhood search direction. In addition, continuing with the example in the above step S304, among the eight new search value arrays corresponding to parameters A, B, E, and F, if the fitness difference value of a certain new search value array corresponding to parameter F is greater than zero and is the largest fitness difference value in this time (i.e., the largest among the eight fitness difference values in this time), then update the freeze countdown value of parameter F from zero to a positive integer value positively correlated with the fitness difference value, while the freeze countdown values of parameters A, B, and E remain zero.

[0100] S307. Determine whether there is any parameter in the set of parameters to be optimized whose current freeze countdown value is zero. If not, decrement the current freeze countdown value of each parameter in the set of parameters to be optimized by 1, and then return to execute step S307. Otherwise, execute step S308.

[0101] In the step S307, that the current freeze countdown values of all parameters in the set of parameters to be optimized are not zero means that it is impossible to return to execute step S304 later. Therefore, they need to be decremented together until the current freeze countdown value of at least one parameter is zero.

[0102] S308. Determine whether there is any fitness difference value greater than zero among the fitness difference values of each of the arrays. If so, update the current optimal search value array to the array corresponding to the minimum fitness among the 2×D″′ new search value arrays, and then execute step S309. Otherwise, directly execute step S309.

[0103] Increment the current iteration count \(t'\) by 1, and determine whether the current iteration count \(t'\) has reached the maximum iteration count \(T\). If so, use the current optimal search value array as the machine-on plan developed for the patient to be put on the machine and the optimal search result for minimizing the incidence of unplanned CRRT machine-off. Otherwise, return to step S304 for execution.

[0104] Based on the foregoing steps S301 - S309, a new heuristic algorithm step is provided to optimize the machine-on plan. During the optimization process, different iteration counts can be temporarily locked for the search results in each best neighborhood search direction based on different fitness difference values, which is conducive to quickly searching for the optimal machine-on plan attribute information. By first quantitatively increasing / decreasing the adjustable parameters step by step and then randomly increasing / decreasing them non-quantitatively, it is also possible to avoid falling into local optimal solutions, which further facilitates quickly and accurately obtaining the optimal search result.

[0105] In addition, when there are multiple nurses to be assigned, the data analysis server is also used to, for each nurse, based on the patient attribute information of the patient to be put on the machine and the nurse attribute information of the corresponding person in the input data, apply the prediction model and combine it with the optimization algorithm to optimize the machine-on plan developed for the patient to be put on the machine, obtain the machine-on plan, the optimal search result for minimizing the incidence of unplanned CRRT machine-off, and then select any optimal search result with the lowest incidence of unplanned CRRT machine-off from the respective optimal search results corresponding to each nurse. Then, when the third incidence of unplanned CRRT machine-off corresponding to this any optimal search result is lower than the first incidence of unplanned CRRT machine-off obtained according to the prediction result, use this any optimal search result as another recommended machine-on plan for the patient to be put on the machine. Finally, transmit the another recommended machine-on plan, the third incidence of unplanned CRRT machine-off, and the nurse attribute information of any nurse corresponding to this any optimal search result to the medical staff workstation device. The medical staff workstation device is also used to output and display the another recommended machine-on plan and the nurse attribute information of any nurse, and determine and output and display another new warning level for unplanned CRRT machine-off based on the third incidence of unplanned CRRT machine-off, so as to provide reference auxiliary information for medical staff when making a decision on whether to put the patient to be put on the machine on CRRT. In this way, the best patient machine-on plan and nurse assignment plan can also be optimized, further enriching the reference information volume.

[0106] In summary, adopting the CRRT unplanned machine-off warning and decision support information system provided in this embodiment has the following technical effects:

[0107] (1) This embodiment provides a new solution for realizing decision support by carrying out risk early warning of unplanned termination of CRRT based on artificial intelligence algorithms, including a medical record database, a data analysis server and a medical staff workstation device. Among them, the medical record database is used to record the medical records of all historical CRRT patients during the machine-on period. The data analysis server is used to calibrate and verify a prediction model for judging whether an unplanned termination of CRRT event occurs based on the successful and failed machine-on medical records and artificial intelligence algorithms, and import the patient attribute information of the patient to be machine-on from the medical staff workstation device, the machine-on plan attribute information of the machine-on plan formulated for the patient to be machine-on, and the caregiver attribute information of the caregiver configured for the patient to be machine-on into the prediction model, obtain a prediction result and feedback it to the medical staff workstation device, so that it can determine the unplanned termination early warning level of CRRT according to the prediction result and output and display it. In this way, the risk of unplanned termination of CRRT can be accurately identified as early as possible, the reference information amount can be enriched, the clinical doctor can be assisted to accurately make CRRT machine-on decisions, and then the incidence rate of unplanned termination of CRRT can be effectively reduced, the CRRT treatment effect can be guaranteed, and it is convenient for practical application and popularization.

[0108] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A CRRT unscheduled machine stop warning and decision support information system, characterized in that, It includes a medical record database, a data analysis server, and a medical staff workstation device. Among them, the data analysis server is communicatively connected to the medical record database and the medical staff workstation device respectively; The medical record database is used to record the medical records of all historical CRRT patients during the machine-on period. Among them, the medical record contains patient attribute information, machine-on plan attribute information, nursing staff attribute information, and the reason for machine-off. The reason for machine-off is divided into planned machine-off and unplanned machine-off; The data analysis server is used to randomly select an equal number of successful machine-on medical records and failed machine-on medical records from the medical record database. Then, negative sample data is sorted out according to the successful machine-on medical records, and positive sample data is sorted out according to the failed machine-on medical records. Finally, all positive sample data and all negative sample data are imported into a machine learning model based on an artificial intelligence algorithm for calibration verification modeling to obtain a prediction model for judging whether an unplanned machine-off event of CRRT occurs. Among them, a successful machine-on medical record refers to a medical record with the reason for machine-off being planned machine-off, and a failed machine-on medical record refers to a medical record with the reason for machine-off being unplanned machine-off; The medical staff workstation device is used to input the patient attribute information of the patient to be put on the machine, the machine-on plan attribute information of the machine-on plan formulated for the patient to be put on the machine, and the nursing staff attribute information of the nursing staff configured for the patient to be put on the machine, and transmit the input data to the data analysis server; The data analysis server is also used to import the input data into the prediction model, output the prediction result, and feedback it to the medical staff workstation device; The medical staff workstation device is also used to determine the early warning level of unplanned machine-off of CRRT according to the prediction result and output and display it, so as to provide reference auxiliary information for medical staff when making a decision on whether to perform CRRT machine-on for the patient to be put on the machine.

2. The CRRT unscheduled machine shutdown warning and decision support information system according to claim 1, wherein The patient attribute information includes the patient's age, the label value of the patient's gender, the patient's disease-related parameters, and / or the patient's coagulation-related parameters. Among them, the patient's disease-related parameters include the label value of the machine-on position, diastolic blood pressure, systolic blood pressure, and / or APACHE II score. The machine-on position is divided into left lateral position, supine position, and right lateral position. The patient's coagulation-related parameters include prothrombin time, activated partial thromboplastin time, fibrinogen, and / or platelet count.

3. The CRRT unscheduled machine shutdown warning and decision support information system according to claim 1, characterized in that, The machine-on plan attribute information includes the planned machine-on duration, the machine-on anticoagulation-related parameters, the machine-on blood purification-related parameters, and / or the machine-on catheter-related parameters. Among them, the machine-on anticoagulation-related parameters include the label value of the number of blood products transfused and / or the type of anticoagulant. The machine-on blood purification-related parameters include the label value of the treatment mode, the label value of the filter type, the pre-dilution fluid volume, the post-dilution fluid volume, the blood flow rate, the arterial pressure, and / or the transmembrane pressure. The machine-on catheter-related parameters include the label value of the catheterization site, the catheterization duration, and / or the label value of the catheter specification.

4. The CRRT unscheduled machine stop warning and decision support information system according to claim 1, wherein The nursing staff attribute information includes the working years of the nursing staff, the label value of the nursing staff's gender, and / or the label value of the nursing staff's position level.

5. The CRRT unscheduled machine shutdown warning and decision support information system according to claim 1, characterized in that Sorting out negative sample data according to the successful machine-on medical records, and sorting out positive sample data according to the failed machine-on medical records, including: Randomly sort all the selected medical records to obtain a sequence of medical records; For each medical record in the medical record sequence, if the corresponding logout reason is planned logout, determine that the occurrence label value of the corresponding CRRT unplanned logout event is zero; otherwise, determine that the occurrence label value of the corresponding CRRT unplanned logout event is one; For each parameter in the medical record other than the logout reason, calculate the correlation coefficient between the corresponding parameter and the CRRT unplanned logout event according to the occurrence label value sequence and the corresponding parameter value sequence that are in one-to-one correspondence with the medical record sequence; Arrange all the parameters in the medical record other than the logout reason in descending order of the correlation coefficient to obtain a parameter sequence; Select the first N parameters from the parameter sequence as key parameters respectively, where N represents a positive integer greater than or equal to 5 and less than or equal to 15; According to the successfully logged-in medical records, organize the negative sample data in the following manner: Use the parameter values of all the key parameters in the successfully logged-in medical records as the first model input item, and use the occurrence label value of the CRRT unplanned logout event corresponding to the successfully logged-in medical records as the first model output item, and then use this first model input item and this first model output item together as the negative sample data; According to the failed logged-in medical records, organize the positive sample data in the following manner: Use the parameter values of all the key parameters in the failed logged-in medical records as the second model input item, and use the occurrence label value of the CRRT unplanned logout event corresponding to the failed logged-in medical records as the second model output item, and then use this second model input item and this second model output item together as the positive sample data.

6. The CRRT unscheduled machine stop warning and decision support information system according to claim 5, characterized in that, For each parameter in the medical record other than the logout reason, calculate the correlation coefficient between the corresponding parameter and the CRRT unplanned logout event according to the occurrence label value sequence and the corresponding parameter value sequence that are in one-to-one correspondence with the medical record sequence, including: For any parameter in the medical record other than the logout reason, extract the corresponding parameter value sequence S1 that is in one-to-one correspondence with the medical record sequence according to all the selected medical records, and perform a normal distribution KS test on the corresponding parameter value sequence S1 to calculate the first test statistic p value p1; Perform a normal distribution KS test on the occurrence label value sequence S2 that is in one-to-one correspondence with the medical record sequence to calculate the second test statistic p value p2; If both the p-value p1 of the first test statistic and the p-value p2 of the second test statistic are greater than the preset threshold, the correlation coefficient r between any of the parameters and the unplanned termination event of CRRT is calculated according to the following formula q : where i represents a positive integer, x i represents the i-th parameter value in S1, y i represents the i-th occurrence tag value in S2, represents the mean of S1, represents the mean of S2.

7. The CRRT unscheduled machine stop warning and decision support information system according to claim 5, wherein After selecting the first N parameters from the parameter sequence as key parameters respectively, import the input data into the prediction model, including: Extract the parameter values of all the key parameters from the input data and import the extraction result into the prediction model.

8. The CRRT unplanned logout warning and decision support information system according to claim 1, characterized in that: The data analysis server is further configured to optimize the onboarding plan formulated for the patient to be onboarded based on patient attribute information of the patient to be onboarded in the input data and nursing attribute information of the nursing staff assigned to the patient to be onboarded, using a prediction model in combination with an optimization algorithm, to obtain an optimized search result for the onboarding plan and for minimizing the incidence of unplanned CRRT offboarding; and when a second unplanned CRRT offboarding rate corresponding to the optimized search result is lower than a first unplanned CRRT offboarding rate obtained based on the prediction result, using the optimized search result as a recommended onboarding plan formulated for the patient to be onboarded; and finally transmitting the recommended onboarding plan and the second unplanned CRRT offboarding rate together to a medical workstation device; The medical workstation equipment is also used to output and display recommended machine-on plans, and to determine and output a new CRRT unplanned discontinuation warning level based on the second CRRT unplanned discontinuation rate, so as to provide reference auxiliary information for medical staff when making decisions on whether to perform CRRT on patients on the machine.

9. The CRRT unscheduled machine stop warning and decision support information system according to claim 8, characterized in that, Based on the patient attribute information of the patient to be connected to the CRRT in the input data and the nursing staff attribute information of the nursing staff assigned to the patient to be connected to the CRRT, a prediction model is applied in combination with an optimization algorithm to optimize the CRRT plan formulated for the patient to be connected to the CRRT, thereby obtaining the optimal search result for the CRRT plan and minimizing the incidence of unplanned CRRT disconnection, including the following steps S301 to S309: S301. Initialize the optimization algorithm parameters including the maximum number of iterations T, randomly generate the initial search value array of the set of parameters to be optimized, and then execute step S302, where the set of parameters to be optimized includes all the parameters in the machine operation plan attribute information of the machine operation plan formulated for the patient to be connected to the machine, and the initial search value array y of the set of parameters to be optimized 0 is expressed as follows: wherein, d″ represents a positive integer less than or equal to D″, and D″ represents the total number of parameters in the set of parameters to be optimized, represents the initial search value corresponding to the d″-th parameter in the set of parameters to be optimized, u c,d″ represents the upper limit of the parameter search space corresponding to the d″-th parameter, l c,d″ represents the lower limit of the parameter search space corresponding to the d″-th parameter, and rand(0, 1) represents a pure decimal random generation function; S302. The patient attribute information of the patient to be admitted and the nursing attribute information of the nursing staff assigned to the patient to be admitted, along with the initial search value array of the parameter set to be optimized, are imported into the prediction model. A first confidence level for determining the occurrence of an unplanned CRRT discharge event is output, and the first confidence level is used as the fitness level corresponding to the initial search value array. Then, step S303 is executed. S303. The initial search value array of the parameter set to be optimized is used as the current optimal search value array, and the current iteration number t′ is initialized to 0. The countdown value of each parameter in the parameter set to be optimized is also initialized to zero, and then step S304 is executed; S304. Based on the current optimal search value array, each parameter in the parameter set to be optimized and whose current prohibited change countdown value is zero is independently incremented and decremented, resulting in 2 × D″′ new search value arrays for the parameter set to be optimized. Then, step S305 is executed, where D″′ represents the total number of parameters in the parameter set to be optimized and whose current prohibited change countdown value is zero, and 1 ≤ D″′ ≤ D″. S305. For each of the 2×D″′ new search value arrays, import the patient attribute information of the patient to be connected to the machine in the input data and the caregiver attribute information of the caregiver configured for the patient to be connected to the machine together with the corresponding array into the prediction model, output the second confidence level for determining the occurrence of an unplanned disconnection event during CRRT, and use this second confidence level as the fitness of the corresponding array, and then execute step S306; S306. For each array, subtract the fitness of the corresponding array from the fitness of the array corresponding to the current optimal search value array to obtain the fitness difference value of the corresponding array. When the fitness difference value is greater than zero and is the largest fitness difference value in this time, update the forbidden change countdown value of the corresponding unique change parameter to a positive integer value positively correlated with this fitness difference value, and then execute step S307, where the unique change parameter refers to the parameter in the set of parameters to be optimized and used to obtain the corresponding array through addition processing or subtraction processing; S307. Determine whether there is any parameter in the set of parameters to be optimized with the current forbidden change countdown value being zero. If not, decrement the current forbidden change countdown value of each parameter in the set of parameters to be optimized by 1, and then return to execute step S307. Otherwise, execute step S308; S308. Determine whether there is any fitness difference value greater than zero among the fitness difference values of each array. If so, update the current optimal search value array to the array in the 2×D″′ new search value arrays corresponding to the minimum fitness, and then execute step S309. Otherwise, directly execute step S309; S309. Increment the current iteration count t′ by 1, and determine whether the current iteration count t′ has reached the maximum iteration count T. If so, use the current optimal search value array as the optimized search result for formulating the connection plan for the patient to be connected to the machine and minimizing the unplanned disconnection rate during CRRT. Otherwise, return to execute step S304.

10. The CRRT unscheduled machine shutdown warning and decision support information system according to claim 1, wherein Determine the early warning level for unplanned disconnection during CRRT according to the prediction result, including: Extract the confidence level for the input data and determined to have an unplanned disconnection event during CRRT according to the prediction result, and use this confidence level as the first unplanned disconnection rate during CRRT; According to all the preset probability ranges corresponding one by one to all the early warning levels for unplanned disconnection during CRRT, if it is found that the first unplanned disconnection rate during CRRT is within the preset probability range corresponding to a certain early warning level for unplanned disconnection during CRRT, then use this certain early warning level for unplanned disconnection during CRRT as the early warning level for unplanned disconnection during CRRT determined according to the prediction result.

Images