JP7629638B2 - Lifestyle intervention system, lifestyle intervention method, program, and program recording medium - Google Patents

Description

The present invention relates to technology that supports appropriate intervention (guidance) for the prevention of lifestyle-related diseases.

Recently, lifestyle-related diseases such as hypertension, diabetes, dyslipidemia, and obesity have become widespread worldwide, and are a major cause of cancer and cardiovascular diseases, which are non-infectious diseases that account for 60% of the causes of death worldwide. Many of these lifestyle-related diseases can be prevented by improving lifestyle habits, and there are high expectations for the therapeutic effects of dietary therapy and exercise therapy in combination with drug therapy after onset. Many of the interventions for lifestyle-related diseases to prevent cardiovascular diseases are based on the results obtained through observational studies, and guidelines have been created by the World Health Organization and related academic societies in each country and implemented as systems. For example, in Japan, health guidance guidelines (Ministry of Health, Labor and Welfare "Standard Health Checkup and Health Guidance Program", Ministry of Health, Labor and Welfare "Dietary Reference Intakes for Japanese People", Ministry of Health, Labor and Welfare "Dietary Balance Guide" by the Ministry of Agriculture, Forestry and Fisheries) have been created, and medical professionals have studied them and put them into practice as health guidance in clinical settings.

In addition, in recent years, in the field of lifestyle-related disease prevention, a device has been proposed that scores the chronological health risks of health checkup recipients and evaluates areas for improvement and areas that do not, and that supports the extraction of targets for preventing the worsening of lifestyle-related diseases (Patent Document 1). Furthermore, a lifestyle-related disease improvement support device is known that uses a regression equation to calculate the time-series correlation between biometric information and lifestyle habits, and uses the obtained correlation to suggest lifestyle habits for the biometric information (Patent Document 2).

JP 2018-160012 A JP 2014-219850 A

However, conventional methods of intervention in lifestyle-related diseases for preventing cardiovascular disease have the following problems. First, the level of evidence obtained from population observational studies is lower than that of randomized controlled trials, and confounding effects due to genetic factors and unobserved factors cannot be completely eliminated, limiting causal inference. Second, even if the results are obtained from randomized controlled trials, the therapeutic effect of the intervention may differ depending on individual characteristics, and this is not taken into consideration. Third, no direct knowledge is available about specific intervention methods. Fourth, even in actual clinical practice, there are time limitations in implementing evidence-based health guidance. Fifth, short-term changes in health behavior have not been evaluated in the past, and the idea of improving lifestyle habits in response to short-term changes has not existed.

The device described in Patent Document 1 aims to make it more efficient to prevent lifestyle-related diseases from worsening based on medical information and health checkup results, but does not compare specific preventive intervention methods. The device described in Patent Document 2 uses a regression equation that expresses correlation to suggest changes in the same type of lifestyle habits that correspond to changes in biological information, and also reflects delay times, but does not appropriately select and suggest one of multiple interventions for biological information.

The present invention has been made in consideration of the above, and provides a lifestyle intervention system, lifestyle intervention method, and program that are capable of selectively presenting appropriate intervention measures for an individual's lifestyle based on machine learning and epidemiological and statistical causal inference.

The lifestyle intervention system of the present invention is characterized by comprising: a storage unit that stores multiple types of intervention content prepared in advance for preventing lifestyle-related diseases in correspondence with parameters; an intervention setting means that, in a time-dependent direction, executes calculations based on an algorithm (machine learning and a probability model) using the corresponding parameters for all interventions, selects the intervention content corresponding to the calculated score, and sets it as the next intervention; a notification unit that notifies the user of the selected intervention content each time an intervention content is selected; a bioinformation acquisition means that receives the user's behavioral record and bioinformation as a result of the notified intervention after notifying the user; and an evaluation means that updates parameters corresponding to the set intervention content as the degree of improvement of lifestyle-related diseases from the acquired behavioral record and bioinformation each time the behavioral record and bioinformation are acquired.

The lifestyle intervention method according to the present invention further comprises a memory unit which stores multiple types of intervention content prepared in advance for preventing lifestyle-related diseases in correspondence with parameters, an intervention setting means which performs calculations based on an algorithm (machine learning and a probability model) using the corresponding parameters for all interventions in a time-dependent direction, selects the intervention content in correspondence with the calculated score, sets it as the next intervention, and notifies the user of the set intervention content via the notification unit, a bioinformation acquisition means which accepts the user's behavioral record and bioinformation as a result of the notified intervention after notifying the user, and an evaluation means which updates the parameters corresponding to the set intervention content as the degree of improvement of the lifestyle-related disease from the acquired behavioral record and bioinformation each time the behavioral record and bioinformation are acquired.

In addition, the program of the present invention is intended to cause a computer to function as a lifestyle intervention system.

In addition, the computer-readable program recording medium of the present invention records a program for causing a computer to function as a lifestyle habit intervention system.

According to these inventions, lifestyle intervention measures are directly and specifically selected and presented to the user, and as a result, the relationship with recorded behavior and biological information is evaluated and parameters are updated, making it possible to provide intervention based on individualized causal effects. Furthermore, it is efficient because continuous and effective intervention measures are designed in accordance with evidence. In addition, because it is possible to utilize technology and data from an information platform capable of real-time monitoring, it is also possible to modify intervention methods through shorter-term feedback compared to the span of health guidance based on conventional health guidance guidelines.

According to the present invention, it is possible to sequentially present intervention strategies based on individualized causal effects on an individual's lifestyle habits.

1 is an overall configuration diagram showing an embodiment of a lifestyle intervention system according to the present invention. 1A and 1B are diagrams showing an embodiment of a personal terminal, in which (A) is an example of a screen in a state where intervention information is presented, and (B) is an example of a screen in a state where biological information has been input. FIG. 13 shows an example of a memory map for storing patient information, etc., where (A) shows attribute information, etc., and (B) shows the intervention strategy (type) and current parameters for the type of intervention. FIG. 13 is a diagram showing an example of a memory map for storing time-series data on interventions and biological information. 13 is a flowchart illustrating an example of a new registration process. 13 is a flowchart illustrating an example of an intervention process.

Figure 1 is an overall configuration diagram showing one embodiment of a lifestyle intervention system according to the present invention. The lifestyle intervention system LIS includes a personal terminal 1 owned by an individual, a server 3 connected to the personal terminal 1 via a network 2 such as the Internet, and a medical institution terminal 4 that can be applied as necessary. The personal terminal 1 may be a desktop personal computer or a mobile terminal such as a smartphone, and is capable of communicating with the server 3 and medical institution terminal 4 connected to the network 2.

The bioinformation measuring device 5 measures the state of the living body, and is adopted according to the purpose. Typically, it is a measuring device such as a blood pressure monitor, a weight scale, a pedometer, and an activity meter. The bioinformation can be obtained using various modes other than the mode obtained from the bioinformation measuring device 5. For example, a meal menu can be input through the operation unit 12 of the personal terminal 1, or a term expressing the degree of physical condition (e.g., very good, good, normal, slightly bad, very bad) can be input, or the doctor's diagnosis can be input in text during medical treatment, or the relevant part can be obtained from the diagnosis result (electronic medical record, etc.) input through the medical institution terminal 4. In addition, for example, when calculating the salt amount from a meal menu, the built-in nutrition calculation software can be applied from the input meal menu, and the salt amount of each food can be calculated using the salt content of each food set in advance to obtain the salt intake for each meal. In addition, a simple test kit (e.g., spot urine test filter paper: for measuring salt in urine) can also be applied.

In this embodiment, the server 3 mainly functions as a memory unit that stores various data. The medical institution terminal 4 is installed in the medical institution and operated by medical staff, and may be one or more units. The medical institution terminal 4 has a memory unit 41 that stores inputted medical information of the patient (such as electronic medical records).

Here, an overview of the lifestyle intervention process using the lifestyle intervention system LIS will be described. Lifestyle intervention is typically performed in the following procedure. First, a person who suffers from or wishes to improve a lifestyle disease such as hypertension and wishes to use this system starts this system (application program) from his/her personal terminal 1 and registers his/her (user) information or attribute information. After registration, the lifestyle intervention system LIS selects one of multiple types of interventions prepared in advance for hypertension at a certain timing and outputs (presents) the content of the intervention to the personal terminal 1. In normal cases, the user performs a lifestyle behavior corresponding to the presented intervention content and inputs bioinformation corresponding to the intervention result within a specified period or at a certain time. The lifestyle intervention system LIS evaluates the intervention based on the content of the intervention and the bioinformation as the intervention result.

Thereafter, the lifestyle intervention system LIS sequentially selects one preferred intervention from the multiple types of interventions using a method described below, outputs (presents) it on the personal terminal 1, and repeats the same procedure as above. If the biometric information has improved to an appropriate level, the system may terminate voluntarily, or if no improvement is seen, a recommendation to see a doctor may be made. Even while using this system, it is also possible to recommend a visit to a medical institution if hypertension, diabetes, dyslipidemia, or obesity is found during a health check. Examples of recommendations to see a doctor include displaying the risk of disease on the personal terminal 1, recommending a visit due to abnormal test results, and recommending a visit by displaying a map image of the relevant medical institution.

The lifestyle intervention system LIS includes an application program that performs intervention processing to prevent lifestyle-related diseases, an information storage unit that stores various information for executing the program, and a control unit that executes the program. Note that the various information is information about all registered users and is stored for each user.

Various configurations of the lifestyle intervention system LIS are possible. For example, as shown in Figure 1, an application program may be downloaded from the server 3 or an external memory (not shown) to the personal terminal 1 at the time of registration and made available for use, while various information is stored in the memory unit 32 of the server 3. In this configuration, the intervention process is performed by the personal terminal 1 exchanging necessary information with the server 3 and using that information to execute the required calculations.

In another embodiment, the application program and various information are stored in a server 3 capable of information processing, and during intervention processing, information from the personal terminal 1 is sent to the server 3, a predetermined calculation is executed on the server 3 side, and the results are sent back from the server 3 and presented on the personal terminal 1. In this embodiment, the personal terminal 1 functions as an input/output unit. Alternatively, the application program and the patient's own information may be imported into the personal terminal 1 at the time of registration, and all processing thereafter may be performed by the personal terminal 1. Alternatively, the memory unit 41 in the medical institution terminal 4 may have the functions of the server 3.

The server 3 comprises a communication processing unit 31 that controls communication with the personal terminal 1 and the medical institution terminal 4, and a memory unit 32. The memory unit 32 comprises a control program memory unit 321, a personal information memory unit 322, an intervention type memory unit 323, an intervention information memory unit 324, and a biological information memory unit 325. The server 3 stores information on all users, while the communication processing unit 31 is adapted to individually exchange information with the personal terminal 1 (its telephone number, email address, or other personally identifiable ID) that is associated with an individual user ID during new registration as described below.

The control program memory unit 321 stores an application program (hereinafter referred to as the control program) for executing the intervention process and each related process.

As shown in FIG. 3A, the personal information storage unit 322 stores the user's state Li and attribute information Xi for each user ID that identifies the user. The state Li (i = 1, 2, ...) is a plurality of states for the user's daily living environment, here the eating environment, for example, L1 "pour soy sauce", L2 "eat fruit", L3 "eat soy products", etc. The types of states are not limited to three, and it is preferable to set an appropriate number and types according to the lifestyle-related disease to be improved. The attribute information Xi includes the user's age, sex, blood pressure value, pulse rate, blood glucose value, weight, body temperature, body mass index (BMI), self-reported stress level, educational history, leisure time, participation in hobby activities, employment type, etc., which indicate the socio-economic situation. The state Li and attribute information Xi may be treated as attribute information. The user ID is set in response to the registration of a name that can identify the user input from the operation unit 12. The registration of the name may not be necessary.

As shown in FIG. 3B, the intervention type storage unit 323 updates and records indices I1, I2, I3, ..., Ii indicating combinations of contents and users prepared in advance as types of interventions (options), and parameters μ1, Σ1, μ2, Σ2, μ3, Σ3, ..., μi, Σi, which are newly calculated for each intervention as described below. The parameters μ and Σ set for each combination of intervention type and user are real values, vectors, matrices, and array variables combining these, and are parameters of a probabilistic model indicating responsiveness based on a probabilistic algorithm that learns the effects of the intervention as outcomes of the user's behavioral records and biological information. For example, if necessary, the responsiveness may follow a linear model, and the mean vector μ and variance matrix Σ of the multidimensional normal distribution that follows the coefficients of the linear model may be used as parameters. In addition to applying the types of interventions prepared in advance as they are, more effective intervention contents may be individually extracted based on the user's information entered at the time of registration.

As shown in Fig. 4, the intervention information storage unit 324 sequentially stores the contents of the selection history of intervention I presented to the user in chronological order via the personal terminal 1. In Fig. 2 (A), an information display area 110 is set on the screen 11a of the display unit 11 of the personal terminal 1, and an example of the intervention content is presented. Known intervention measures for preventing lifestyle-related diseases include diet-related measures, exercise-related measures, sleep-related measures, drinking/smoking-related measures, and stress management measures, and in this embodiment, any one or more of these can be adopted.

As shown in Fig. 4, the biometric information storage unit 325 sequentially stores biometric information D acquired in chronological order from the biometric information measuring device 5 or the operation unit 12, or input from the medical institution terminal 4 regarding medical treatment results. Fig. 2 (B) shows an example of biometric information (numerical portion) input by a user in the information display area 110 of the screen 11a of the display unit 11.

In this way, in FIG. 4, for example, for a time series t1, t2, t3, ..., tn, interventions I are input in the order of I2(t1), I4(t2), I3(t3), ..., Ii(t4), and correspondingly acquired biometric information D is input in the order of D(t1), D(t2), D(t3), ..., D(tn).

Furthermore, the personal terminal 1 is equipped with a control unit 10, typically consisting of a processor (CPU), a display unit 11, an operation unit 12, and a memory unit 13. The display unit 11 displays images, presents intervention content, displays input bioinformation, etc. The part of the display unit 11 that functions as an output unit for the intervention process may output audio in addition to images, and in the case of audio, it may be equipped with a speaker. The operation unit 12 may be a transparent piezoelectric element panel that is laminated on the display unit 11 and functions as a touch panel, and inputs information via tapping, flicking, etc., to various buttons including character key buttons displayed on the display unit 11. It may also include a physical operation unit. It may also include an audio operation unit.

The control unit 10 reads out the control program into a main memory (not shown) and executes it, thereby functioning as a registration unit 101, an intervention setting unit 102, a biometric information acquisition unit 103, and an evaluation unit 104. The control unit 10 also functions as a timer 105 that performs timing operations.

The registration unit 101 executes new user registration processing required when using this system. In the new registration processing, a control program returned from the server 3 is downloaded in response to an application download request from the personal terminal 1. The registration unit 101 also starts the control program, and accepts input of the type of lifestyle-related disease to be improved, information such as a name for identifying the user, status information Li, and attribute information Xi via the operation unit 12, and stores the information in the memory unit 13 as shown in Figures 3A and 3B. If necessary, the acquired information Li and Xi are sent to the server 3 for storage in the personal information memory unit 322 or the like.

The intervention setting unit 102 selects the intervention content at a certain timing and presents it to the user via the display unit 11. More specifically, the intervention setting unit 102 selects one of the multiple interventions for the next intervention based on the previous intervention content and the intervention result, such as acquired biological information. When there are multiple interventions i=1, 2, ..., K aiming at a health outcome (here, the outcome of lifestyle habits due to the intervention: biological information, etc.), this algorithm evaluates the health outcome {Xi} of the intervention result while performing each intervention, thereby:
Score y = (ΣXi) / 2Ni(t) + √(logT / 2Ni(t)) ... (Equation 1)
Here, the sum of the first term in (Equation 1) is over the selected period of intervention i. In addition, this intervention content (intervention option) decision algorithm does not deterministically select the intervention i(t) that maximizes the score y, but may be an algorithm that selects each intervention option by a probability weighted by its parameter μ(i).

As shown in the above (Equation 1), the score y is obtained by adding a correction value (second term) based on the intervention time T and the number of interventions Ni(t) up to that point to the first term, which is the average of the observed value of the intervention effect i(t). This method is an upper confidence bound (UCB) strategy as an example of a bandit algorithm in machine learning. That is, the score y is a value obtained by adding a correction term to the sample mean, and the value of the correction term is set to be larger for interventions with fewer selections Ni(t). Therefore, an intervention with a small sample mean but a small number of selections may be selected. In this way, when an intervention with a large sample mean is simply selected, "utilization" is performed, and when an intervention with a small sample mean but a small number of selections is selected, "exploration" is performed. In this way, in this embodiment, the UCB strategy is used to repeatedly select interventions while balancing exploration and utilization, aiming to maximize the target for improvement. In addition, by performing evaluation using the minimum number of tests necessary for evaluating effectiveness, interventions with a low score y, i.e., interventions that do not produce results, may be terminated in a short period of time, i.e., removed from the intervention target.

The algorithm used in the intervention setting unit 102 may use a method of Logistic Thompson Sampling, which learns a model for predicting outcomes from attributes for each intervention option by logistic regression, and updates hyperparameters representing the mean and variance of the logistic regression coefficients, taking into account uncertainty according to the number of samples. In this case, when selecting an intervention option, the option with the highest linear prediction value of the logistic regression using the coefficients randomly selected from the distribution represented by the updated hyperparameters and the user attributes is selected. In the case of Logistic Thompson Sampling, the regret upper bound as described above is not known, but it is empirically known that the regret can be kept low. The hyperparameters μik, Σik representing the distribution of the parameter βik for each selection value i and the attribute subscript k are stored. When selecting an intervention, the linear predictor score y = Σβik × Xk ... (Equation 2) is calculated using the coefficient βik randomly selected from the multidimensional normal distribution N (μik, Σik) represented by the hyperparameters.
Select the i for which is maximum.

The algorithm used in the intervention setting unit 102 may be a model incorporating random effects that consider the individual level, option level, and further interaction between the individual and option levels in the above-mentioned Logistic Thompson Sampling. When updating the hyperparameters after each intervention, it is assumed that the parameters of each coefficient follow a normal distribution, and the mean and variance are used as hyperparameters, and updating is performed using Laplace approximation. Hyperparameters μijk, Σijk that represent the distribution of parameters βijk for each selection value i, individual identification number j, and attribute subscript k are stored. When selecting an intervention, a linear predictor score y=Σβijk×Xk...(Equation 3) is calculated for individual identification number j using coefficient βijk randomly selected from the multidimensional normal distribution N(μijk, Σijk) expressed by the above hyperparameters.
Select the i for which is maximum.

Furthermore, as an outcome, for example an annual health check or consultation, it is possible to additionally apply a method of aggregating interventions and their results for a group and updating them using batch processing.

The intervention setting unit 102 may include options such as "eat less miso soup," "don't put soy sauce on the table," "eat less pickles," "eat more fruits," and "eat more vegetables" as examples of methods for intervening in hypertension. The types of interventions may range from a few to several hundred or more. Furthermore, the intervention setting unit 102 can notify the user of the intervention information by e-mail or by using, for example, a "messaging service." The "messaging service" is an extension of the well-known short message service (SMS) in the personal terminal 1, and refers to a function for sending and receiving multimedia messages including images and videos. Furthermore, when implemented in an app, it may be performed by push notification. Push notification is a function in which a message notified to an app is displayed on the personal terminal even when the user has not started the personal terminal or app.

The intervention setting unit 102 can also use the timer 105 to set the frequency and timing of intervention as appropriate or individually, and may also automatically vary the frequency and timing by observing the results of the intervention (for example, the span may be extended as the improvement increases). The timing of intervention can be, for example, by time of day, or early morning, morning, noon, afternoon, before bedtime, etc., and the frequency can be every meal, every day, once a week, every two weeks, once a month, or for a longer span. The optimal timing for each of these can be selected as an option, and optimization can be performed by an algorithm executed by the intervention setting unit 102.

The biometric information acquisition unit 103 imports biometric information D(t) measured by a biometric measuring device 5 or the like. The imported biometric information D(t) includes the measured value and the measurement date and time. The evaluation unit 104 evaluates the improvement due to the lifestyle intervention based on the imported time-series biometric information D(t). The improvement can be evaluated, for example, based on the degree of improvement over time and the difference from the normal value. Furthermore, the evaluation unit 104 calculates new parameters based on the parameters of the intervention item and the biometric information that is the result of the intervention each time an intervention is performed, and updates the parameters of the corresponding intervention item in FIG. 3(B). By sequentially going through such parameter update processing, the type of intervention selected at the next intervention timing is optimized.

Figure 5 is a flow chart showing an example of a new registration process. First, it is determined whether or not a new registration is requested (step S1). If there is no request for new registration, this flow is exited. On the other hand, if there is a request for new registration, and if a control program, which is an application program, is stored in the server 3, it is downloaded from the server 3 to the personal terminal 1 (step S3), and then the control program is started (step S5). Next, a screen for accepting input information of the name, status, and attribute information is displayed, and the necessary information is accepted through input on the screen (step S7). The input information is stored in the memory unit 13 (step S9), and, if necessary, is sent to the server 3 side and stored in the memory unit 32.

Figure 6 is a flowchart showing an example of an intervention process. Initialization at start-up is performed following new registration processing, and involves selecting the lifestyle-related disease to be improved and setting the content of the corresponding intervention (step S11). The intervention content may be prepared in advance according to the type of lifestyle-related disease. In addition, the state L may also be input after selecting the lifestyle-related disease to be improved. Furthermore, the initialization may be performed by initially setting a general or empirical score for each intervention according to the type of intervention.

After the initialization is completed, the timer 105 determines whether the first or next intervention timing has been reached (step S13). If it is determined that it is time to intervene, the intervention setting unit 102 performs the calculation process of the score y shown in the above (Equation 1) to (Equation 3) for all intervention items, assuming a hypothetical case in which the intervention is selected, and determines the intervention when the calculated value becomes the maximum value as the current intervention content. When one intervention content is selected from a plurality of intervention contents in this way, the intervention setting unit 102 displays the selected intervention content on the display unit 11, for example, as shown in FIG. 2(A) (step S15). The first intervention may be a previously set or general one, or may be determined with a random element included.

Next, it is determined whether or not biometric information such as blood pressure values has been accepted (step S17). It is preferable that the acceptance of biometric information is specified as the acquisition timing at the time of initial registration or initialization. When the biometric information is input, an evaluation process is performed on the biometric information acquired as a result of the intervention (step S19). Then, the evaluation result is obtained and recorded in the memory unit 13 in an updated manner (step S21). Alternatively, it may be recorded in chronological order. It is preferable that the evaluation result is displayed on the display unit 11. In addition, for this intervention, a new score is calculated based on the current score for the intervention type and the biometric information that is the result of the intervention, and the score for the corresponding intervention type in Figure 3 (B) is updated.

Next, a suggestion is made as to whether or not to terminate the lifestyle intervention program (step S23). If the program is to be terminated, for example, information during the intervention period is stored on the server 3 side and made available for future use. If the symptoms do not improve or even worsen, it is preferable to be notified of a recommendation to see a doctor. On the other hand, if the program is to be continued, the intervention setting unit 102 moves on to a process of determining the next intervention timing. Note that the decision as to whether or not to terminate the intervention program is preferably made by treating it as an automatic extension if there is no response from the user within a specified period, for example by the time of the next intervention.

In addition, multiple ways of conveying the intervention content may be prepared so that the user can follow the intervention content as much as possible while engaging in lifestyle activities. For example, when using voice, it is conceivable to prepare separate items for messages using different tones of voice as separate types of intervention.

The evaluation items after each intervention may be aimed not only at health behavior or input values of biological information, but also at the onset of disease, assuming a longer-term effect. Below, we will provide an overview of the calculation method for each value, including this case.

<1> State S is expressed by a set (D, L, X) of D, which indicates the presence or absence of a disease, L, which indicates lifestyle habits, and X, which indicates personal attributes.
・D: Represents diseases such as high blood pressure, diabetes, myocardial infarction, stroke, and cancer.
・State L = {Li} is a lifestyle, for example, L1: "put soy sauce on", L2: "eat fruit", L3: "eat soy products"
Attribute information X={Xi}: Represents individual attributes such as age, sex, body mass index (BMI), etc.

<2> Intervention I = {Ii}: intervention for lifestyle change. For example, possible options are I0: "do nothing", I1: "provide guidance for improving lifestyle habits", I2: "suggest alternative methods (such as presenting a different seasoning method for those who use soy sauce)", and I3: "present options for improvement and ask the patient to choose".

<3>・reward function: g(D,L,X,I)=-P(D(t+1)|S(t))
Here, the measured biometric information of each user is combined as a reward function to evaluate the risk score of cardiovascular disease, and the sign of the risk score is inverted and used to select the intervention. The option that maximizes the reward from the intervention, i.e., the intervention option that minimizes the risk of disease, is efficiently selected.
・P(D(t+1) | S(t)) is defined as follows: if D(t)=1, then it has the value 1, otherwise it has the value h(S(t)), where h(S(t)) is the regression equation.

h(S(t)) is the risk score of the above disease, which is reflected in the reward function. D(t)=1 is the onset of the disease, which corresponds to, for example, the actual onset of a stroke. In this case, the patient is no longer covered by this system and is treated at a medical institution.

<4>・Possibility of transition: P _T (L(t+1) |S(t),I(t))
・Factorization assumption: P _T (L(t+1) |S(t),I(t))
=P(L(t+1) |L(t),X(t))×f(X(t),I(t))×α ₀
The possibility of transition is the probability of transition, and is the probability distribution function that expresses the probability of transitioning to a new lifestyle L(t+1) when intervention I(t) is administered to a subject with certain lifestyle habits and attributes S(t), and is expressed as _PT . This part is the formula required when confirming the effectiveness of this system through simulation.

The factorization assumption is the assumption that the above transition probability is decomposed, and the transition probability is expressed as the product of P(L(t+1)|L(t), X(t)), which is the probability representing natural changes in lifestyle, and f(X(t),I(t)), which is the coefficient of the effect of intervention I(t) on a person with attribute information X(t). In addition, α ₀ is the normalization constant for the probability.

・P(L(t+1) |L(t),X(t)): By regression equation or machine learning.
f(X(t),I(t))∈[0.5,2.0]
In the above equation, the coefficient of responsiveness to the intervention can take values between 0.5 and 2.0.

Here, the responsiveness coefficient refers to the responsiveness of certain attribute information X(t) to the intervention options required in the simulation of the transition probability above.

<5>・Target function: Maximize the survival function R(T)＝E[Π ^T _t=1 (1―D(t))].

The target function is the overall objective function, which we aim to maximize. The formula represents the probability of living from period t=1 to period t=T without contracting the disease, and the goal is to avoid contracting the disease, i.e., successful prevention.

As described above, the lifestyle intervention system of the present invention is characterized by comprising a memory unit that stores multiple types of intervention content prepared in advance for preventing lifestyle-related diseases in correspondence with parameters, an intervention setting means that performs calculations based on an algorithm (machine learning and probability model) using the corresponding parameters for all interventions in the time direction, selects the intervention content corresponding to the calculated score, and sets it as the next intervention, a notification unit that notifies the user of the set intervention content each time the intervention content is set, a bioinformation acquisition means that accepts the user's behavioral record and bioinformation as a result of the notified intervention after notifying the user, and an evaluation means that updates parameters corresponding to the set intervention content as the degree of improvement of lifestyle-related diseases from the acquired behavioral record and bioinformation each time the behavioral record and bioinformation are acquired.

The lifestyle intervention method according to the present invention further comprises a memory unit which stores multiple types of intervention content prepared in advance for preventing lifestyle-related diseases in correspondence with parameters, an intervention setting means which executes a calculation based on an algorithm (machine learning and a probability model) using the corresponding parameters for all interventions in a time-dependent direction, selects an intervention content corresponding to the calculated score, sets it as the next intervention, and notifies the user of the set intervention content via the notification unit, a bioinformation acquisition means which accepts the user's behavioral record and bioinformation as a result of the notified intervention after notifying the user, and an evaluation means which updates the parameters corresponding to the set intervention content as the degree of improvement of the lifestyle-related disease from the acquired behavioral record and bioinformation each time the behavioral record and bioinformation are acquired.

In addition, the program of the present invention is intended to cause a computer to function as a lifestyle intervention system.

In addition, the computer-readable program recording medium of the present invention records a program for causing a computer to function as a lifestyle habit intervention system.

According to these inventions, lifestyle intervention measures are directly and specifically selected and presented to the user, and the resulting recorded behaviors (e.g., diet, exercise, number of steps, sleep time, mood, perceived stress, etc.) and correlations with biological information are evaluated and parameters are updated, making it possible to provide intervention based on individualized causal effects. Furthermore, it is efficient because continuous and effective intervention measures are designed in accordance with evidence. In addition, because it is possible to utilize technology and data from an information platform capable of real-time monitoring, it is also possible to modify intervention methods through shorter-term feedback compared to the span of health guidance based on conventional health guidance guidelines.

In addition, it is preferable that the storage unit stores the types of interventions for each user. This makes it possible to prepare and set appropriate interventions according to the user's illness.

It is also preferable that the notification unit is a display unit that displays an image. This makes it possible to utilize the display unit provided on the terminal as is.

It is also preferable that the intervention setting means is capable of setting at least one of the frequency and timing of intervention. This allows for manual or automatic adjustments based on observation of the results of the intervention (e.g., the span can be lengthened as improvement is observed).

In addition, the lifestyle habit intervention system according to the present invention preferably comprises a server and a terminal operated by a user, which are communicatively connected via a network, and the terminal is preferably configured to download a control program from the server and start the control program when a new user registers to use the system. With this configuration, the terminal is able to execute a lifestyle habit intervention process by downloading the control program from the server.

LIS Lifestyle intervention system 1 Personal terminal 11 Display unit (notification unit)
12 Operation unit (biometric information acquisition means)
10 Control unit 101 Registration unit 102 Intervention setting unit (intervention setting means)
103 Biometric information acquisition unit (biometric information acquisition means)
104 Evaluation unit (evaluation means)
105 Timer 2 Network 3 Server 32 Storage unit 321 Control program storage unit 5 Biometric information measuring device (biometric information acquiring unit)

Claims (8)

a storage unit that stores a plurality of types of interventions prepared in advance for preventing lifestyle-related diseases in association with the parameters;
An intervention setting means for performing a calculation based on an algorithm using the corresponding parameters for all interventions at each intervention timing , selecting one preferred intervention from the calculated score, and setting the selected intervention as the next intervention;
a notification unit that notifies a user of the set intervention ;
a biometric information acquiring means for accepting biometric information of the user as a result of the notified intervention after the notification to the user;
and an evaluation means for updating a parameter corresponding to the set intervention as a degree of improvement of a lifestyle-related disease based on the set intervention and the biological information acquired as a result of the set intervention, each time the biological information is acquired , and storing the updated parameter in the memory unit . 2. The lifestyle intervention system according to claim 1, wherein the storage unit stores types of interventions for each user. The lifestyle intervention system according to claim 1 or 2, wherein the notification unit is a display unit that displays an image. The lifestyle intervention system according to any one of claims 1 to 3, wherein the intervention setting means is capable of setting at least one of the frequency and timing of intervention. A server and a terminal operated by a user are connected to each other via a network so as to be capable of communicating with each other,
3. The lifestyle intervention system according to claim 1, wherein the terminal comprises the notification unit and a registration unit that downloads a control program from the server and starts the control program when a new user registers to use the system. a storage unit that stores a plurality of types of interventions prepared in advance for preventing lifestyle-related diseases in association with parameters, and an intervention setting means executes a calculation based on an algorithm using the corresponding parameters for all interventions at each intervention timing , selects one preferable intervention from the calculated score, and sets it as the next intervention;
The notification unit notifies the user of the set intervention ,
a biometric information acquiring means for accepting the biometric information of the user as a result of the notified intervention after notifying the user;
the evaluation means updates a parameter corresponding to the set intervention as a degree of improvement of the lifestyle-related disease based on the set intervention and the bioinformation acquired as a result of the set intervention each time the bioinformation is acquired , and stores the updated parameter in the memory unit . A program for causing a computer to function as a lifestyle intervention system according to any one of claims 1 to 4. A computer-readable program recording medium having recorded thereon a program for causing a computer to function as a lifestyle habit intervention system according to any one of claims 1 to 4.

Images