WO2023175702A1

WO2023175702A1 - Prognosis management support system and prognosis management support method

Info

Publication number: WO2023175702A1
Application number: PCT/JP2022/011496
Authority: WO
Inventors: 裕太郎井山; 将司末崎; 正典篠原
Original assignee: 株式会社日立製作所
Priority date: 2022-03-15
Filing date: 2022-03-15
Publication date: 2023-09-21

Abstract

A recommendation model building unit 102 creates an individual recommendation model in which prognostic management data for a target patient is used to update weightings for predictor variables in a multivariate analysis equation of a standard recommendation model corresponding to the disease or disorder of the target patient, a recommendation calculation unit 103 uses the standard recommendation model or the individual recommendations model to calculate recommendations regarding prognostic behavior for the target patient from the prognostic management data of the target patient, and, during the prognostic period of the target patient, the recommendation calculation unit calculates recommendations that use the individual recommendation model when an individual recommendation model has been created, and the recommendation model building unit uses the prognostic management data of the target patient that was used in calculating the recommendations to update the weightings of the predictor variables in the multivariate analysis equation of the individual recommendation model.

Description

Prognosis management support system and prognosis management support method

The present invention relates to a prognosis management support system and a prognosis management support method.

Rehabilitation has traditionally been used to restore physical function. Regenerative medicine is a treatment method that rebuilds destroyed body tissues by transplanting cells that have been harvested and processed from humans. It is possible to apply regenerative medicine to reconstruct neural networks, which have traditionally been considered difficult to regenerate, and this may also restore functions lost due to damage to the central nervous system, etc. is being born. However, transplantation does not immediately cure the disease; in order to regenerate tissue and restore function, it is necessary to properly manage the prognosis over several months to several years. In other words, during the prognosis period, patients need to continue appropriate medication and rehabilitation.

Patent Document 1 discloses a rehabilitation support system for preventing a decrease in motivation for rehabilitation. Specifically, the past patient most similar to the target patient is extracted, and the extracted past patient's observation history information is used to predict and output the target patient's future observation information.

JP2016-197330A

In regenerative medicine, tissue regeneration and functional recovery after transplantation hold the key to the success or failure of treatment, and individual patient differences have a large impact on the degree of tissue regeneration and functional recovery. Therefore, there is no guarantee that the optimal medication or rehabilitation for one patient will also be optimal for other patients. For this reason, currently, patient prognosis management has to rely on the experience and intuition of doctors and therapists.

For this reason, decision-making regarding prognosis management in clinical settings largely depends on the knowledge and experience of individual physicians and therapists in charge, and tends to become person-dependent. Regenerative medicine is a cutting-edge medical field that is still actively researched and tested, so there are cases in which there is not enough evidence to begin with. Therefore, even if cells are transplanted, optimal prognosis management may not be carried out, and the expected therapeutic effect may not be obtained.

For this reason, we output recommendations for optimal medication and rehabilitation to doctors and patients based on past patient medication and rehabilitation knowledge and data on target patients, maximizing treatment effects for patients' diseases and disorders. It is desirable to aim for this.

A prognosis management support system that is an embodiment of the present invention is a prognosis management support system that supports prognosis actions including medication and rehabilitation of a target patient during a prognosis period, and includes:
A standard recommendation model database constructed for each disease or disorder and registered with standard recommendation models defined by multivariate analysis formulas;
comprising a recommendation model construction unit and a recommendation calculation unit, and an analysis unit that calculates recommendations regarding prognostic actions for the target patient from the target patient's prognosis management data;
The recommendation model construction unit creates an individual recommendation model in which weights for explanatory variables in the multivariate analysis formula of the standard recommendation model corresponding to the target patient's disease or disorder are updated using the target patient's prognosis management data,
The recommendation calculation unit calculates recommendations regarding prognostic actions for the target patient from the target patient's prognosis management data using a standard recommendation model or an individual recommendation model,
During the target patient's prognosis period, the recommendation calculation unit calculates a recommendation using the individual recommendation model if an individual recommendation model has been created, and the recommendation model construction unit calculates the prognosis of the target patient used to calculate the recommendation. Update the weighting of explanatory variables in the multivariate analysis formula of the individual recommendation model using management data.

Outputs prognostic action recommendations that link medication and rehabilitation that are optimized for each patient, maximizing the effectiveness of treatment for a patient's disease or disorder.

Other objects and novel features will become apparent from the description of this specification and the accompanying drawings.

FIG. 1 is a diagram showing an overall image of a prognosis management support system. FIG. 2 is a functional block diagram of a prognosis management support system. FIG. 3 is a diagram showing the hardware configuration of an analysis section. This is the overall processing flow for building a recommendation model. This is the construction flow of a standard recommendation model. This is a recommendation output flow. This is the data structure of the patient database. This is the data structure of the medical database. This is the data structure of the prognosis status database. This is the data structure of the prognosis status database. This is the data structure of the prognosis status database. This is the data structure of the prognostic behavior database. This is the data structure of the prognostic behavior database. This is the data structure of the case database. This is the data structure of the recommendation model database. This is an example of a display screen displayed to a doctor. This is an example of an automatic alert for a patient using a wearable sensor. It is a figure which shows the modification of a prognosis management support system.

FIG. 1 shows the overall picture of the prognosis management support system of this embodiment, and FIG. 2 shows a functional block diagram of the prognosis management support system. The prognosis management support system 100 is a system that quantitatively evaluates the patient's condition during the prognosis period and outputs prognostic action recommendations that are optimized for each patient according to the patient's condition and the degree of tissue regeneration and functional recovery. be. In this example, patients who need to continue taking medication and rehabilitation (collectively referred to as prognosis actions) for a certain period of time for reasons such as restoring the function of organs treated with regenerative medicine etc. Prognosis management refers to optimizing prognosis actions for patients during the period during which a doctor examines the patient (referred to as the prognosis period). Although the treatment content is not limited to regenerative medicine, in the case of regenerative medicine, there are large individual differences in the progress of tissue regeneration and functional recovery, and it is possible to achieve high effectiveness by proposing medication instructions and rehabilitation instructions that are optimally related to the individual patient. You can expect it.

The prognosis management support system 100 will be explained with reference to FIGS. 1 and 2. Basic information and medication information about the target patient are registered in the patient database 121 as patient information. Hereinafter, the database may be referred to as DB. FIG. 7 shows the data structure of the patient DB 121. Basic information and medication information are registered in the patient DB 121 for each patient ID that uniquely identifies a patient whose prognosis is to be managed (target patient). Basic information includes information such as the patient's age, gender, height, weight, disease history, and allergies. Medication information records the medication history and includes information such as drug name, administration period, and side effects.

Medical information on examinations, diagnoses, and treatments by doctors for target patients who require prognosis management support by this system is registered in the medical care database 122. FIG. 8 shows the data structure of the medical treatment DB 122. In the medical treatment DB 122, medical examination information, diagnosis information, and treatment information are registered for each medical treatment ID that uniquely identifies medical treatment by a doctor. Here, by linking the medical care ID with the patient ID, it is possible to specify which patient the registered contents of the medical care ID correspond to. The medical examination information includes the medical examination method used by the doctor, the medical examination results, and the like. Diagnostic information includes the name of the disease diagnosed by the doctor, its condition, cause, etc. The treatment information includes implementation details and schedules of treatments (surgeries, etc.) that require prognosis management.

During the prognosis period, prognostic state information is registered in the prognostic state database 123 and prognostic action information is registered in the prognostic action database 124. Here, the prognosis status information indicates the physical condition of the patient, and the prognosis behavior information indicates the content of the patient's medication and rehabilitation. The prognosis state DB 123 and the prognosis action DB 124 are collectively referred to as a prognosis database 125. Some of the inputs to the prognosis DB 125 are manually input by a doctor, therapist, or patient, and some are automatically input from a sensing device.

FIGS. 9A, 9B, and 10 show the data structure of the prognosis status DB 123. 9A to 9B are examples of data tables in which objective evaluation values of prognostic conditions are registered, and FIG. 10 is an example of a data table in which subjective evaluation values of prognostic conditions are registered. The objective evaluation value is an evaluation of the patient's condition based on data quantified by the sensor. For example, by having a patient wear a wearable sensor, data regarding the patient's biological information can be collected non-invasively and continuously. In the prognosis state DB 123a shown in FIG. 9A, the values of the digital biomarkers and wearable sensing data collected in this way, and the evaluation based thereon are registered. The prognosis state DB 123b shown in FIG. 9B is another example of an objective evaluation value, and is obtained by quantifying and evaluating the patient's motion using 3D motion capture. In the prognosis state DB 123b, 3D motion capture values and evaluations based on them are registered. FIG. 10 shows an example in which ADL (Activities of Daily Living) evaluation and ePRO (electronic Patient Reported Outcome) are registered as examples of subjective evaluation values of prognostic conditions. The subjective evaluation value is an evaluation based on the senses of the observer, and although it varies depending on the individuality of the observer, it includes useful information that cannot be evaluated by a sensor. ADL evaluation is an evaluation of whether a person is able to perform the minimum daily activities necessary for daily life, and the observation results (evaluation scores) by the therapist are registered here. ePRO is a self-evaluation report directly input by the patient using a smartphone or the like.

FIGS. 11A and 11B show the data structure of the prognostic action DB 124. FIG. 11A is an example of a data table in which medication information is registered among prognostic actions, and FIG. 11B is an example of a data table in which rehabilitation information is registered among prognostic actions. Medication information includes the type/name of the drug, amount, date and time of administration, side effects, etc. Rehabilitation information includes the type/name of rehabilitation, intensity, implementation date and time, special notes at the time of rehabilitation, etc.

Data in the prognosis DB 125 is registered for each prognosis ID that uniquely identifies a prognosis state or prognosis action. By linking the prognosis ID with the patient ID and the medical treatment ID, it is possible to understand which treatment of which patient the status of the prognosis period registered in the prognosis DB 125 is based on.

Patient information (patient data) from the patient DB 121 (see Fig. 7), medical information (medical data) from the medical treatment DB 122 (see Fig. 8), and prognosis DB 125 (Figs. 9A, B, Fig. 10, Fig. 11A, B). Prognosis information (prognosis data) is input to the analysis section 101 via the data input section 120. The recommendation calculation unit 103 of the analysis unit 101 uses the recommendation model to output prognostic actions (positive recommendations) recommended to the patient, and the recommendation model construction unit 102 creates a recommendation model based on the information used to output the recommendation. By updating , we build an individual recommendation model that makes optimized recommendations for patients. By continuously learning the individual recommendation model during the prognosis period, it becomes possible to make recommendations optimized for the patient. In the following, when separately indicating whether a recommendation model is optimized for a patient, a recommendation model trained using patient's individual information will be referred to as an "individual recommendation model," and a recommendation model trained using patient's individual information will be referred to as an "individual recommendation model," and a recommendation model that has not been trained using patient's individual information will be referred to as "individual recommendation model." The model is called the standard recommendation model.

On the other hand, similar case information of patients who have cases similar to the case of the target patient is stored in the case database 111. The doctor extracts similar case information (similar case data) from the case DB 111 and sets prognostic actions (negative recommendations) that are prohibited for the patient. Note that it is possible to continuously accumulate evidence in the case DB 111 by adding data of target patients whose prognosis has been managed using this system after the prognosis management is completed.

The recommendation output unit 130 integrates the positive recommendation and the negative recommendation and outputs a recommendation for the patient. As described later, the recommendation destination may be a doctor or a target patient. Further, the output recommendations are simultaneously recorded in the recommendation database 131.

FIG. 3 shows the hardware configuration of the analysis section 101. The analysis unit 101 is an information processing device such as a server. In the information processing device, a CPU (processor) 201, a main storage device 202, an auxiliary storage device 203, an I/O interface 204, a network interface 205, and the like are communicably connected via a bus 206. An input device 207 such as a keyboard or a pointing device, and an output device 208 such as a display or printer are connected to the I/O interface 204. Furthermore, it becomes possible to connect to an external network via the network interface 205. The function of the analysis unit 101 is that a program stored in an auxiliary storage device 203 is read into the main storage device 202 by a CPU (processor) 201 and executed, thereby performing a predetermined process in cooperation with other hardware. This will be realized. A program executed by the information processing device, its function, or means for realizing the function are expressed as functional blocks such as the recommendation model construction unit 102 and the recommendation calculation unit 103 in FIG.

FIG. 4 shows the overall processing flow of recommendation model construction in the prognosis management support system 100. The premise of this flow is that since the recommendation model in the prognosis management support system 100 is a statistical model, a certain amount of patient information for prognosis management has already been accumulated through demonstration experiments and trials, and based on that accumulation, the recommendation model is a statistical model. A standard recommendation model corresponding to the disease or disorder has already been created and registered in the recommendation model database 110. FIG. 13 shows the data structure of the recommendation model DB 110. In the recommended model DB 110, recommended model information such as analytical formulas is registered for each model ID that uniquely identifies a recommended model.

The doctor inputs the target patient's disease or disorder name into the analysis unit 101 (S01). The analysis unit 101 searches the recommendation model DB 110 and checks whether there is a standard recommendation model for the input disease or disorder name (S02). If it does not exist, a standard recommendation model is constructed (S03) and registered in the recommendation model DB 110. The method for constructing the standard recommendation model will be described later.

If a standard recommendation model exists or after the standard recommendation model is constructed, the data input unit 120 inputs patient data, medical care data, and prognosis data regarding the target patient to the analysis unit 101. The data input unit 120 not only inputs these data into the analysis unit 101 using a data input instruction from a doctor or the like as a trigger, but also inputs these data into the analysis unit 101 using a database update as a trigger or at a predetermined time. It is desirable to set it so that it is automatically entered.

The recommendation calculation unit 103 calculates proactive recommendations by substituting input data into a standard recommendation model. The positive recommendation output by the standard recommendation model is a recommendation that recommends a prognostic action in which medication and rehabilitation are optimally associated for a standard patient, based on cases accumulated in the prognosis management support system 100.

The analysis unit 101 displays the calculated prognostic action recommendation to the doctor (S06). At this time, if the target patient is wearing a wearable sensor, the doctor can set so that an alert (recommendation for the target patient) is automatically sent as necessary.

Thereafter, the prognosis data for the target patient is accumulated in the prognosis database 125, thereby obtaining a track record for proactive recommendations. The data input unit 120 inputs or automatically inputs prognosis data corresponding to the results of active recommendations into the analysis unit 101.

The recommendation model construction unit 102 checks whether an individual recommendation model for the target patient already exists (S08), and if it does not exist, registers an individual recommendation model for the target patient based on the standard recommendation model ( S09). Then, the individual recommendation model is updated using the active recommendation and the track record for the active recommendation as learning data (S10). The framework of the individual recommendation model is the same as that of the standard recommendation model, but the weighting is optimized for the target patient. For example, the standard recommendation model is a multivariate analysis formula such as F=a ₁ x ₁ + a ₂ x ₂ +...+a _n x _n + b, where the objective variable is F and the explanatory variable x _i (1≦i≦n). When expressed, the objective variables and explanatory variables are the same in the standard recommendation model and the individual recommendation model, but in the individual recommendation model, the coefficients a _i (1≦i≦n) and b in the multivariate analysis has been modified by learning based on prognostic data.

After that, calculation of active recommendations using the individual recommendation model and updating of the individual recommendation model will be repeated until the target patient finishes using the solution of this system. That is, after inputting the first prognosis data, the calculation of the active recommendation in step S05 is executed using the individual recommendation model. Thereby, it is possible to make prognostic action recommendations optimized for each target patient.

The target patient's use of the solution of this system ends at the end of the prognosis period (S11). Thereafter, necessary anonymization processing is performed on the series of data of the target patient, and the standard recommendation model is updated as learning data (S12). This is to update the standard recommendation model based on the state of tissue regeneration and functional recovery at the end of the prognosis period.

FIG. 5 shows the construction flow of the standard recommendation model by the recommendation model construction unit 102 in step S03. A standard recommendation model is created for each disease or disorder. Further, here, a case where a recommendation model is created using a multivariate analysis formula will be explained as an example.

First, target variable candidates are selected (S21-S22). For example, it is desirable to select objective data (sensing data) that reflects the disease status, tissue regeneration, and degree of functional recovery from prognostic data as the objective variable. For example, the amount of dopamine secretion is selected as a candidate variable of interest.

Next, explanatory variable candidates are selected (S23-S24). As explanatory variables, parameter factors that can influence the selected objective variable candidates are selected from patient data, medical care data, prognostic state data, and prognostic behavior data. For example, candidate explanatory variables include gender, age, blood type, height, weight, ADL score, pulse rate, blood pressure, heart rate, 3D motion capture data, type, timing, and amount of medication (time axis and intensity for each medication). ) Type, timing, and intensity of rehabilitation (time axis and intensity for each set).

Check whether there is a sufficient number of data sets to perform multivariate analysis (S25), and if there is a sufficient number of data sets, create a temporary multivariate analysis formula with undetermined coefficient values (S26). . Multivariate analysis generally requires at least 10 times as many data sets as the number of explanatory variables. The data set is input into a temporary multivariate analysis formula, and standard partial regression coefficients are determined (S27). In order to verify whether the multivariate analysis formula obtained in this way can accurately estimate the target variable, statistical indicators such as the coefficient of determination are calculated (S28), and a predetermined accuracy index value is calculated. It is determined whether the above conditions are met (S29). If the obtained multivariate analysis formula is greater than or equal to the accuracy index value, it is determined as a standard recommendation model (S30), and if the obtained multivariate analysis formula is less than the accuracy index value, the objective variable and explanatory variable are Start over with selection.

The standard recommendation model obtained in this way uses sensing data that reflects the target patient's condition (for example, dopamine secretion level) as the objective variable, and sets explanatory variables as input variables for the recommendation model (input explanatory variables) and output from the recommendation model. This is a multivariate analysis formula in which the output explanatory variable is the output explanatory variable. Input explanatory variables are appropriately selected variables that indicate the characteristics of the target patient obtained from patient data, the disease or disorder obtained from clinical data, the characteristics of the treatment content, the patient's prognostic state and prognostic behavior obtained from prognostic data. The output explanatory variable is a variable indicating the content of the patient's prognosis action that is output as a positive recommendation.

FIG. 6 shows the flow for outputting recommendations to target patients.

First, decide on a negative recommendation for the target patient. For this reason, the case DB 111 is used. FIG. 12 shows the data structure of the case DB 111. In the case DB 111, case information is registered for each case ID that uniquely identifies a case. The analysis unit 101 collates the patient data and medical care data of the target patient with the case data stored in the case DB 111, and extracts similar case data of the target patient (S41). The analysis unit 101 outputs candidates for prohibited acts from the extracted similar case data (S42). For example, from the case data shown in FIG. 12, since a side effect of "dizziness during rehabilitation during high blood pressure" is observed, an action such as performing rehabilitation during high blood pressure can be output as a prohibited action.

The doctor sets prohibited actions and, if necessary, sets an automatic alert on the target patient's wearable sensor (S43). For example, a doctor should prohibit rehabilitation if the pulse exceeds 140 beats per minute, or if the systolic blood pressure during exercise increases by 40 mmHG or more, or if the diastolic blood pressure increases by 20 mmHG or more during exercise. If the target patient is wearing a wearable sensor, an alarm will be generated to the target patient to stop rehabilitation if the value of sensing data from the wearable sensor meets these conditions. Set. This makes it possible to ensure compliance with prohibited acts even when the patient is undergoing rehabilitation at home. Note that prohibited acts (negative recommendations) are not expected to change significantly unless the patient's condition changes significantly. Therefore, the setting of prohibited acts may be reviewed at an appropriate timing based on the doctor's judgment.

Next, proactive recommendations for the target patient are determined. Therefore, the data input unit 120 inputs the target patient's data set defined as an input explanatory variable in the recommendation model to the analysis unit 101 (S44). The recommendation calculation unit 103 calculates a positive recommendation using the recommendation model (S45).

For example, as described above, the recommendation model uses sensing data that reflects the condition of the target patient (for example, dopamine secretion level) as the objective variable, and uses the explanatory variables as input variables to the recommendation model and output from the recommendation model. Assume that it is a multivariate analytical formula with variables as follows. The data of the target patient is input to the input explanatory variable, and an output explanatory variable that makes the target variable a normal value is calculated. The output explanatory variable is a variable that indicates the content of the patient's prognosis action. Specifically, prognostic actions such as the type, timing, and amount of medication (time axis and intensity for each drug), and rehabilitation type, timing, and intensity (time axis and intensity for each set) are output as the content of the recommendation.

FIG. 14 shows a GUI screen of the prognosis management support system 100. The GUI screen 300 is a screen displayed to the doctor, and display contents can be selected using tabs 301. Here, the "recommendation tab" is shown open. A summary of the patient's condition is displayed based on patient data, medical care data, and prognosis data, and recommendations made by the analysis unit 101 are also displayed. Active recommendations are displayed as recommended actions, and negative recommendations are displayed as prohibited actions.

FIG. 15 shows an example of an automatic alert for a patient using a wearable sensor. If a patient is wearing a wearable sensor to measure their physical condition, it is preferable that the wearable sensor generates an alarm, but automatic alerts can be sent to a device that the patient carries at all times, such as a smartwatch or smartphone. may be set. As shown in FIG. 15, notifications can be made in various situations.

FIG. 16 shows a modification of the prognosis management support system. In the analysis unit 101 described above, the recommendation model construction unit 102 constructs a standard recommendation model, whereas the prognosis management support system 100b uses a standard recommendation model construction unit 402 that exclusively constructs a standard recommendation model. It has a model construction section 401. In fact, it is desirable to collect a large number of cases in order to construct an accurate standard recommendation model from cases that include medically unknown content. Therefore, in the prognosis management support system 100b, a standard model construction unit 401 creates a standard recommendation model, and a plurality of analysis units 501i (1<i≦m) calculate recommendations for target patients and create individual recommendation models. . For example, the analysis unit 501i is installed in a server installed at each medical institution, and the standard model construction unit 401 is installed in a server different from the analysis unit 501i.

The analysis unit 501i performs the same processing as the analysis unit 101 except for creating and updating the standard recommendation model. Note that the prognosis management database 512 is a collective term for the patient DB 121, the medical treatment DB 122, and the prognosis DB 125. Prognosis management data of patients whose prognosis period has ended is anonymized and then provided to the standard model construction unit 401. The standard recommendation model construction unit 402 creates and updates a standard recommendation model using anonymized prognosis management data collected from each analysis unit 501i stored in the prognosis management database 412. Further, the standard model construction unit 401 extracts case data from the prognosis management data accumulated in the prognosis management DB 412 and accumulates it in the case DB 411.

100, 100b: Prognosis management support system, 101: Analysis unit, 102: Recommendation model construction unit, 103: Recommendation calculation unit, 110: Recommendation model database, 111: Case database, 120: Data input unit, 121: Patient database, 122 : Medical treatment database, 123: Prognosis state database, 124: Prognosis behavior database, 125: Prognosis database, 130: Recommendation output unit, 131: Recommendation database, 201: CPU, 202: Main storage device, 203: Auxiliary storage device, 204: I/O interface, 205: network interface, 206: bus, 207: input device, 208: output device, 300: GUI screen, 301: tab, 401: standard model construction section, 402: standard recommendation model construction section, 410: Standard recommendation model database, 411: case database, 412: prognosis management database, 501: analysis section, 502: individual recommendation model construction section, 510: individual recommendation model database, 512: prognosis management database.

Claims

A prognosis management support system that supports prognosis actions including medication and rehabilitation of a target patient during the prognosis period,
A standard recommendation model database constructed for each disease or disorder and registered with standard recommendation models defined by multivariate analysis formulas;
comprising a recommendation model construction unit and a recommendation calculation unit, and an analysis unit that calculates a recommendation regarding a prognostic action for the target patient from the prognosis management data of the target patient;
The recommendation model construction unit creates an individual recommendation model in which weights for explanatory variables in a multivariate analysis formula of a standard recommendation model corresponding to a disease or disorder of the target patient are updated using prognosis management data of the target patient;
The recommendation calculation unit calculates a recommendation regarding a prognostic action for the target patient from the prognosis management data of the target patient using a standard recommendation model or an individual recommendation model,
During the prognosis period of the target patient, if the individual recommendation model has been created, the recommendation calculation unit calculates a recommendation using the individual recommendation model, and the recommendation model construction unit calculates a recommendation using the individual recommendation model. A prognosis management support system that updates weighting for explanatory variables in a multivariate analysis formula of the individual recommendation model using prognosis management data of the target patient.
In claim 1,
The prognosis management data includes patient data indicating the characteristics of the target patient, medical data indicating the disease or disorder of the target patient and characteristics of its treatment, prognostic status data and prognosis indicating the physical condition of the target patient during the prognosis period. including prognostic behavior data indicating the prognostic behavior of the target patient during the period;
The objective variable of the multivariate analysis formula of the standard recommendation model is selected from the prognosis status data, the explanatory variables of the multivariate analysis formula of the standard recommendation model include an input explanatory variable and an output explanatory variable, and the input explanatory variable The variables are selected from the patient data, the clinical data, the prognostic status data, and the prognostic behavior data, and the output explanatory variables are selected from the prognostic behavior data;
The recommendation calculation unit inputs the prognosis management data of the target patient into an input explanatory variable of a multivariate analysis formula of the standard recommendation model or the individual recommendation model, and calculates an output explanatory variable that makes the objective variable a normal value. Prognosis management support system.
In claim 2,
The output explanatory variables of the multivariate analysis formula of the standard recommendation model or the individual recommendation model include the type, time axis, and intensity of medication, and the type, time axis, and intensity of rehabilitation.
In claim 1,
We have a case database that registers past patients' diseases or disabilities, medication records, and rehabilitation records.
The analysis unit extracts similar case data of the target patient by comparing the case database with prognosis management data of the target patient,
A prognosis management support system in which prohibited acts of the target patient are set based on similar case data of the target patient.
In claim 1,
The target patient wears a wearable sensor for measuring the physical condition,
A prognosis management support system in which automatic alerts based on recommendations calculated by the analysis unit can be set in the wearable sensor.
In claim 1,
A prognosis management database that anonymizes and accumulates prognosis management data of a patient whose prognosis action has been supported by the prognosis management support system after the prognosis period ends;
A prognosis management support system comprising: a standard recommendation model construction unit that constructs the standard recommendation model using prognosis management data accumulated in the prognosis management database.
In claim 6,
A prognosis management support system in which the standard recommendation model construction unit and the analysis unit are implemented on different servers.
A prognosis management support method using a prognosis management support system that supports prognosis actions including medication and rehabilitation of a target patient during a prognosis period, the method comprising:
The prognosis management support system is constructed for each disease or disorder and includes a standard recommendation model database in which a standard recommendation model defined by a multivariate analysis formula is registered, a recommendation model construction section, and a recommendation calculation section, an analysis unit that calculates recommendations regarding prognostic actions for the target patient from the patient's prognosis management data;
The recommendation model construction unit creates an individual recommendation model in which weights for explanatory variables in a multivariate analysis formula of a standard recommendation model corresponding to a disease or disorder of the target patient are updated using prognosis management data of the target patient;
The recommendation calculation unit calculates a recommendation regarding a prognostic action for the target patient from the prognosis management data of the target patient using a standard recommendation model or an individual recommendation model,
During the prognosis period of the target patient, if the individual recommendation model has been created, the recommendation calculation unit calculates a recommendation using the individual recommendation model, and the recommendation model construction unit calculates a recommendation using the individual recommendation model. A prognosis management support method that updates weights for explanatory variables in a multivariate analysis formula of the individual recommendation model using prognosis management data of the target patient.
In claim 8,
The prognosis management data includes patient data indicating the characteristics of the target patient, medical data indicating the disease or disorder of the target patient and characteristics of its treatment, prognostic status data and prognosis indicating the physical condition of the target patient during the prognosis period. including prognostic behavior data indicating the prognostic behavior of the target patient during the period;
The objective variable of the multivariate analysis formula of the standard recommendation model is selected from the prognosis status data, the explanatory variables of the multivariate analysis formula of the standard recommendation model include an input explanatory variable and an output explanatory variable, and the input explanatory variable The variables are selected from the patient data, the clinical data, the prognostic status data, and the prognostic behavior data, and the output explanatory variables are selected from the prognostic behavior data;
The recommendation calculation unit inputs the prognosis management data of the target patient into an input explanatory variable of a multivariate analysis formula of the standard recommendation model or the individual recommendation model, and calculates an output explanatory variable that makes the objective variable a normal value. Prognosis management support method.
In claim 9,
The output explanatory variables of the multivariate analysis formula of the standard recommendation model or the individual recommendation model include the type, time axis, and intensity of medication, and the type, time axis, and intensity of rehabilitation.
In claim 8,
The prognosis management support system has a case database in which past patients' diseases or disabilities, medication records, and rehabilitation records are registered,
The analysis unit extracts similar case data of the target patient by comparing the case database with prognosis management data of the target patient,
A prognosis management support method in which prohibited acts for the target patient are set based on similar case data of the target patient.
In claim 8,
The target patient wears a wearable sensor for measuring the physical condition,
A prognosis management support method in which an automatic alert based on a recommendation calculated by the analysis unit can be set in the wearable sensor.
In claim 8,
The prognosis management support system includes a prognosis management database that anonymizes and accumulates the prognosis management data of the patient after the prognosis period of the patient whose prognosis behavior has been supported by the prognosis management support system, and a standard that constructs the standard recommendation model. It has a recommendation model construction department,
In the prognosis management support method, the standard recommendation model construction unit constructs the standard recommendation model using prognosis management data accumulated in the prognosis management database.
In claim 13,
A prognosis management support method in which the standard recommendation model construction unit and the analysis unit are implemented on a different server.