WO2013161191A1 - 保健指導対象者選定条件作成支援装置 - Google Patents
保健指導対象者選定条件作成支援装置 Download PDFInfo
- Publication number
- WO2013161191A1 WO2013161191A1 PCT/JP2013/002301 JP2013002301W WO2013161191A1 WO 2013161191 A1 WO2013161191 A1 WO 2013161191A1 JP 2013002301 W JP2013002301 W JP 2013002301W WO 2013161191 A1 WO2013161191 A1 WO 2013161191A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- health
- health guidance
- data
- guidance target
- target person
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
Definitions
- the present invention relates to a health guidance business for an insurer, and more particularly to a device that supports the work of an insurer that determines conditions for selecting a health guidance target person.
- the insurer is providing health guidance for health promotion of the insured.
- a certain standard is set for the health check result of the insured person, and health guidance is given to a person (applicable person) that satisfies the standard.
- the standard is set so that a person who needs to improve his / her habits or a person who has a high risk of future sickness is applicable.
- non-applicable people it is determined that there is no need to improve their lifestyle habits and health guidance is not provided.
- the problem is dealing with new people who meet the health guidance standards.
- those who do not meet the health guidance standards for the year there are those who meet the health guidance standards in the following year and become eligible persons. Therefore, even if the insurer gives health guidance to the relevant person in the relevant year and realizes the health promotion of the relevant person, if the number of new persons increases every year, the insured as a whole Cannot achieve health promotion.
- the insurer In order for the insured to improve the health of the insured, it is desirable for the insurer to take measures to promote the health of all the insured. However, since it costs enormous costs, the insurer is independent from the health guidance standards. In this year, select the insured person corresponding to the next year from the insured person of the current year, and for the selected limited group, the relevant year Wants to provide health guidance.
- Patent Document 1 An example of a technique for selecting a health guidance subject is described in Patent Document 1.
- the technique described in Patent Document 1 is a function for outputting a predicted value of a medical cost reduction effect when an insurer sets a selection condition for a target person and gives guidance to a health guidance target group that satisfies the condition. Is provided. By using this support system, the insurer can know the medical cost reduction effect when a specific selection condition is set.
- the object of the present invention is the problem as described above, that is, it is difficult to create a condition for selecting a health guidance target person from the viewpoint of whether or not there is a possibility of meeting the health guidance standard in the next year. It is to provide a health guidance target person selection condition creation support device that solves the problem.
- the health guidance target person selection condition creation support device Health guidance criteria predetermined by first health check data which is personal check data of the first period and health check data of the second period which is the next period after the first period.
- a plurality of health examination items of the health examination data are defined as a plurality of explanatory variables, expressed by a polynomial composed of the explanatory variables and a coefficient for each explanatory variable, and the individual is the health examination data of the second period.
- the health guidance target person selection condition creation support method is: Health guidance criteria predetermined by first health check data which is personal check data of the first period and health check data of the second period which is the next period after the first period.
- a health guidance target person selection condition creation support method executed by a device that includes a memory that stores a label value that indicates whether or not the condition is satisfied, and a processor connected to the memory,
- the processor is
- a plurality of health examination items of the health examination data are defined as a plurality of explanatory variables, expressed by a polynomial composed of the explanatory variables and a coefficient for each explanatory variable, and the individual is the health examination data of the second period.
- a discrimination model for discriminating whether or not the health guidance standard is met is learned using the first health examination data and the label value, Generating a combination of the plurality of health diagnosis items as the plurality of explanatory variables and the value of the coefficient as the plurality of explanatory variables in the discrimination model after the learning as a health guidance target person selection condition; The structure is adopted.
- the insurer can create a condition for selecting a health guidance target person from the viewpoint of whether or not there is a possibility of meeting the health guidance standard in the next period.
- the health guidance target person selection condition creation support device 1 is a condition for selecting a health guidance target person from the viewpoint of whether or not there is a possibility that the health guidance standard is met. Has the function to create.
- the health guidance target person selection condition creation support apparatus 1 includes a communication interface unit (communication I / F unit) 11, an operation input unit 12, a screen display unit 13, a storage unit 14, and a processor 15 as hardware.
- the communication I / F unit 11 includes a dedicated data communication circuit and has a function of performing data communication with various devices (not shown) connected via a communication line (not shown).
- the operation input unit 12 includes an operation input device such as a keyboard and a mouse, and has a function of detecting an operator operation and outputting the operation to the processor 15.
- the screen display unit 13 includes a screen display device such as an LCD or a PDP, and has a function of displaying various information such as operation menus and selection conditions on the screen in accordance with instructions from the processor 15.
- the storage unit 14 includes a storage device such as a hard disk or a semiconductor memory, and has a function of storing processing information and a program 14P necessary for various processes in the processor 15.
- the program 14P is a program that realizes various processing units by being read and executed by the processor 15, and is read by an external device (not shown) or a computer via a data input / output function such as the communication I / F unit 11. It is read in advance from a possible storage medium (not shown) and stored in the storage unit 14.
- Main processing information stored in the storage unit 14 includes health check data 14A, a flag 14B, an insurer's desired condition 14C, a discrimination model 14D, and a selection condition 14E.
- Health checkup data 14A is personal checkup data for a certain past year (hereinafter referred to as a reference year).
- the medical checkup data 14A is divided by individual.
- the health check-up data for one individual for one year consists of a personal ID that uniquely identifies the individual, personal information such as the year of medical examination, age of medical examination, sex, etc., height, weight, waist circumference, BMI value, minimum blood pressure, maximum blood pressure, It has various test values such as blood glucose level and neutral fat, and various inquiry results such as whether or not “the amount of drinking is 500 ml or more per day”.
- one year is one period, but the period is arbitrary as long as it is less than one year.
- the flag 14B is a label value for each individual indicating whether or not a predetermined health guidance standard is met in the medical examination in the year following the base year.
- the predetermined health guidance standard is, for example, a health guidance standard focusing on built-in fat type obesity.
- the present invention is not limited to such an example.
- the insurer's desired condition 14C is a condition regarding an individual that the insurer wants to actively participate in health guidance. For example, a condition that an individual in their 40s wants to actively participate in health guidance, or an individual who satisfies the result of a test item in a specific health examination meets a certain condition actively participates in health guidance. It is a condition that wants.
- the discriminant model 14D is a model that shows the relationship between the personal health check data and whether or not the individual meets a predetermined health guidance standard in the next year of the base year.
- the discriminant model 14D may be a linear regression model, a logistic regression model, or the like.
- the discrimination model 14D is generally a polynomial composed of a plurality of explanatory variables and their coefficients (parameters). As individual explanatory variables, individual health check items in the health check data are used. All health check items in the health check data may be explanatory variables, or some health check items in the health check data may be set as explanatory variables.
- the health checkup data corresponds to age of visit, sex, height, weight, waist circumference, BMI value, diastolic blood pressure, diastolic blood pressure, blood sugar level, triglyceride, “drinking amount is 500 ml or more per day” If there are a total of 11 health checkup items with the results of the interviews, all 11 may be used as explanatory variables. For example, only 8 health checkup items excluding age, sex and height may be used as explanatory variables. good.
- the selection condition 14E is a condition for selecting a health guidance target person created from the discriminant model 14C after learning.
- the selection condition 14E includes a combination of health check items and their coefficients, and a determination threshold value.
- the combination of the health check item and its coefficient means a combination of a coefficient whose value is not 0 and a health check item as an explanatory variable corresponding to the coefficient among a plurality of coefficients in the discriminant model 14D after learning.
- the determination threshold value can be determined in the discriminant model 14D after learning that a certain individual has a probability that the person corresponds to the health guidance standard in the next year of the reference year is equal to or greater than a preset threshold value (for example, 1/2). , Means the minimum value of the total value of the coefficients included in the combination.
- the processor 15 has a microprocessor such as a CPU and its peripheral circuits, and reads and executes the program 14P from the storage unit 14, thereby causing the hardware and the program 14P to cooperate to implement various processing units. have.
- main processing units realized by the processor 15 there are an input unit 15A, a discriminant model learning unit 15B, and a condition creating unit 15C.
- the input unit 15A inputs the health check data 14A, the flag 14B, the insurer's desired condition 14C, and the discrimination model 14D before learning from the communication I / F unit 11 or the operation input unit 12, and stores them in the storage unit 14. It has the function to do.
- the discrimination model learning unit 15B reads the health check data 14A, the flag 14B, the insurer's desired condition 14C, and the discriminant model 14D before learning from the storage unit 14, and the health check data 14A, the flag 14B, and the insurer's desired condition 14C. Is used to learn the discriminant model 14D and to store the discriminated model 14D after learning in the storage unit 14.
- the discriminant model learning unit 15B uses, as a positive example, the individual's health diagnosis data corresponding to the health guidance standard in the next fiscal year in the health diagnosis data 14A, and corresponds to the health guidance standard in the next year.
- the medical examination data of the individuals who did not exist are used as negative examples.
- the discriminant model learning unit 15B determines whether a term that represents the likelihood of the discriminant model, a penalty term that depends on the number of coefficients whose values are not 0, and an individual who does not satisfy the insurer's desired conditions.
- the coefficient of the discriminant model 14D is learned so as to optimize an objective function having a penalty term depending on meeting the health guidance standard in the year.
- the condition creating unit 15C reads the discriminant model 14D after learning from the storage unit 14, the combination of a coefficient whose value in the discriminated model 14D after learning is not 0 and the health check item as an explanatory variable corresponding to the coefficient,
- the determination threshold value is generated as the selection condition 14E and stored in the storage unit 14.
- the condition creating unit 15 ⁇ / b> C has a function of reading the selection condition 14 ⁇ / b> D from the storage unit 14 and outputting it to the screen display unit 13 or outputting it to the outside through the communication I / F unit 11.
- the input unit 15A inputs the health check data 14A, the flag 14B, the insurer's desired condition 14C, and the discrimination model 14D before learning from the communication I / F unit 11 or the operation input unit 12, and stores them in the storage unit 14. (Step S1).
- the discrimination model learning unit 15B reads the health diagnosis data 14A, the flag 14B, the insurer's desired condition 14C, and the discrimination model 14D before learning from the storage unit 14, and the health diagnosis data 14A, the flag 14B, and the insurer.
- the discriminant model 14D is learned using the desired condition 14C (step S2). That is, the discriminant model learning unit 15B uses, as a positive example, the health check data of the individual who met the health guidance standard in the next fiscal year in the health check data 14A, and the health of the individual who did not meet the health guidance standard in the next year.
- the coefficient value of each explanatory variable in the discriminant model 14D for appropriately relating the individual's health diagnosis data and whether or not the individual meets the predetermined health guidance standard in the next year The insurer's desired condition 14C is learned as much as possible.
- the discriminant model 14D after learning is stored in the storage unit 14.
- condition creation unit 15C reads the discriminant model 14D after learning from the storage unit 14, and among the plurality of explanatory variables of the discriminant model 14D after learning, determines that the coefficient value is a combination of explanatory variables other than zero.
- the threshold value is stored in the storage unit 14 as the selection condition 14E, is output to the screen display unit 13, or is output to the outside through the communication I / F unit 11 (step S3).
- a B is expressed as A_B.
- the superscript is expressed with a hat.
- a B is written as A ⁇ B.
- Step S1 The input unit 15A inputs the health diagnosis data 14A, the flag 14B, the insurer's desired condition 14C, and the discrimination model 14D before learning.
- N represents the number of individuals who are candidates for health guidance.
- X_n is health check data for the reference year of individual n.
- M is the number of health examination items.
- the input unit 15A binarizes the health check data as preprocessing.
- the threshold value of each inspection item may use the health guidance criteria set by the Ministry of Health, Labor and Welfare. For example, the BMI value is binarized to 1 if 25 or more, and to 0 otherwise. Note that this process is not necessary when the input health checkup data is already binarized.
- Step S2 The value obtained by the discrimination model 14D is the following P.
- the discriminant model learning unit 15B learns the parameter using the discriminant model 14D for calculating this value.
- a logistic regression model that can output a probability that a certain X_n is a positive example or a negative example may be used.
- the mathematical structure of logistic regression will be described.
- X is an M-dimensional explanatory variable corresponding to the health examination data of the base year
- W is an M-dimensional weight vector
- the logistic regression model is expressed by the following equation.
- P (Y 1
- X; W) 1 / (1 + exp (W ⁇ ⁇ T ⁇ X))... (1)
- P (Y 0
- ⁇ ; ⁇ ) represents the conditional probability of ⁇ when ⁇ is given as a parameter and ⁇ is given.
- a superscript T represents transposition of a vector.
- L (W) can be maximized by a method according to the gradient method.
- the value of the parameter W that maximizes L (W) is W *.
- Figure 3 shows an example of health check items corresponding to the value of W * that maximizes the objective function given by Equation 3 above.
- the example shown in FIG. 3 shows the result of obtaining the coefficient value of each explanatory variable that maximizes the objective function with the five health check items, ie, waist circumference, BMI, blood sugar, lipid, and not drinking, as explanatory variables. It is an example.
- the coefficients of all five explanatory variables have values other than zero.
- the number of condition items of the selection condition is 5 items.
- the insurer in order for an insured person to participate in health guidance, it is necessary for the insurer to disclose the reason for selection to the insured person and to gain an understanding of the insured person. If a complicated selection reason with a large number of condition items is set, it becomes difficult to obtain the insured's understanding. Therefore, it is desirable that the number of condition items as selection conditions is as small as possible. Further, the objective function of the above three equations cannot be optimized along the insurer's desired condition 14C.
- the discriminant model learning unit 15B learns W * that optimizes the objective function according to the following four expressions instead of the objective function according to the above three expressions.
- is the norm of W, and norm 1 is used.
- F is an N-dimensional vector, and the nth element is W ⁇ ⁇ T ⁇ X_n.
- L ′ is a normalized graph Laplacian. ⁇ and ⁇ are parameters for adjusting the balance of the first term, the second term, and the third term on the right side.
- Equation 4 The first term on the right side of Equation 4 above is the same as the right side of Equation 3 above and represents the likelihood of the discriminant model.
- the second term on the right side of the above four formulas is a penalty term that depends on the number of coefficients whose values are not zero.
- the second term on the right side has an effect of reducing the number of explanatory variables in which the coefficient of each explanatory variable of the discriminant model, that is, the element of W * is not zero.
- the third term on the right side of the above four formulas is a penalty term that depends on an individual who does not meet the insurer's desired conditions falling under the health guidance standards.
- the third term on the right side has the effect of adjusting the weight of each element of W * so as to meet the conditions desired by the insurer.
- W * is learned so that the insurer's desired condition is satisfied as much as possible. Therefore, the non-zero component of W * after learning is extracted and the condition is extracted. Then, it is possible to generate health guidance target person selection conditions that are less in the number of condition items and that meet the conditions desired by the insurer.
- the graph Laplacian is an N ⁇ N matrix.
- N represents the number of individuals who are candidates for health guidance.
- the normalized graph Laplacian is the same as the graph Laplacian in this example.
- Each row and column in the graph Laplacian and normalized graph Laplacian in FIG. 4 represents an individual.
- the first line in FIG. 4 indicates that there is a link between person 1 and person 2.
- the second line in FIG. 4 indicates that there is a link between person 1 and person 2.
- the third line in FIG. 4 indicates that there is no link.
- the insurer's desired condition 14C is a condition that an individual in their 40s wants to actively participate in health guidance, for example, there is a link that satisfies the condition desired by the insurer. Indicates. In the example of FIG. 4, individuals 1 and 2 meet the conditions desired by the insurer (individuals in their 40s), and individual 3 does not meet the conditions desired by the insurer (individuals in their 40s) Is shown.
- the normalized graph Laplacian L ′ itself may be stored in the storage unit 14 as the insurer's desired condition 14C and applied to the objective function of Equation 4.
- the condition itself that an individual in their 40s wants to actively participate in health guidance is stored in the storage unit 14 as an insurer's desired condition 14C, and the normalized graph Laplacian L ′ is calculated from the condition. It may be generated and applied to the objective function of Equation 4.
- W * is an M-dimensional vector, and among the elements of the M-dimensional vector, the value of the non-zero element and the health check item corresponding to the element are stored as a selection condition 14E in the storage unit 14. That is, (W * _j, health examination item j) ⁇ j
- Figure 5 shows an example of health check items corresponding to the value of W * that maximizes the objective function given by Equation 4 above.
- the coefficient value of each explanatory variable that maximizes the objective function according to the equation 4 having five health check items such as abdominal circumference, BMI, blood glucose, lipid, and no drinking as explanatory variables is obtained. It is a result.
- the value of the coefficient of BMI and blood glucose level is zero.
- FIG. 6 shows an example of the selection condition 14E generated by extracting the health check item corresponding to the non-zero element and the coefficient value corresponding to the health check item from FIG.
- the number of condition items of the selection condition is reduced from 5 items to 3 items.
- the coefficient value is treated as a score.
- the value of W ⁇ ⁇ T ⁇ X when P (Y 1
- TH is 0.5, but other values may be used.
- the condition creating unit 15C includes the selection condition 14E generated as described above, the condition (W * _j, health diagnosis item j) ⁇ j
- the insurer can create a condition for selecting a health guidance target person from the viewpoint of whether or not there is a possibility that the health guidance standard will be met in the next period.
- the reason is that a plurality of health check items of the health check data are set as a plurality of explanatory variables, expressed by a polynomial composed of this explanatory variable and a coefficient for each explanatory variable.
- the discrimination model 14D for discriminating whether or not the data meets the health guidance standards is determined based on the health check data 14A in the reference year and whether each individual meets the health guidance standards in the health check data in the next year of the reference year. This is because the learning is performed using the flag 14B indicating such, and a combination of a plurality of health check items and coefficient values as a plurality of explanatory variables in the discriminant model 14D after learning is generated as the selection condition 14E.
- the learning of the discriminant model 14D learns the value of the coefficient of the discriminant model 14D so as to optimize the objective function according to Equation 4 having a penalty term that depends on the number of coefficients whose values are not zero.
- the discriminant model 14D is optimized so as to optimize the objective function according to Equation 4 having a penalty term that depends on a person who does not satisfy the condition desired by the insurer being in compliance with the health guidance standard. This is because the value of the coefficient is learned.
- a threshold value for determining whether or not an individual meets a predetermined standard in the next year can be generated as a part of the selection condition 14E.
- X; W) TH in the above equation 1 is calculated and stored in the storage unit 14 as part of the selection condition 14E. It is to do.
- the health guidance target person selection condition creation support device 2 is a condition for selecting a health guidance target person from the viewpoint of whether or not there is a possibility that it meets the health guidance standards. And a function of selecting a health guidance subject according to the created selection conditions.
- the health guidance target person selection condition creation support device 2 includes a communication interface unit (communication I / F unit) 21, an operation input unit 22, a screen display unit 23, a storage unit 24, and a processor 25 as hardware.
- the communication I / F unit 21, the operation input unit 22, and the screen display unit 23 have the same functions as the communication I / F unit 11, the operation input unit 12, and the screen display unit 13 in the first embodiment. Yes.
- the storage unit 24 includes a storage device such as a hard disk or a semiconductor memory, and has a function of storing processing information and programs 24P necessary for various processes in the processor 25.
- the program 24P is a program that realizes various processing units by being read and executed by the processor 25, and can be read by an external device (not shown) or a computer via a data input / output function such as the communication I / F unit 21. It is read in advance from a possible storage medium (not shown) and stored in the storage unit 24.
- health diagnosis data 24A, a flag 24B, an insurer's desired condition 24C, a discrimination model 24D, a selection condition 24E, and a selection health diagnosis data 24F There is a selector 24G.
- the health diagnosis data 24A, the flag 24B, the insurer's desired condition 24C, the discrimination model 24D, and the selection condition 24E are the health diagnosis data 14A, the flag 14B, the insurer's desired condition 14C, the discrimination model 14D in the first embodiment. The same as the selection condition 14E.
- the health check data 24F for selection is personal health check data in a year (hereinafter referred to as a selection year) in which a health instructor is selected.
- the medical checkup data 24F is divided by individual.
- the health check data for one individual in the current year has the same items as the health check data 24A of the reference year. That is, the health check data 24F of one individual includes personal ID that uniquely identifies the individual, personal information such as the year of medical examination, age of medical examination, sex, etc., height, weight, waist circumference, minimum blood pressure, maximum blood pressure, blood sugar level, medium It has various test values such as sex fat, and various inquiry results such as whether or not “the amount of drinking is 500 ml or more per day”.
- the health check data 24F for selection may consist only of health check data of individuals not corresponding to the prescribed health guidance standards, or health check data of individuals who do not meet the prescribed health guidance standards and individuals who do not meet the criteria. Data may be mixed.
- the selector 24G is information for identifying an individual selected as a health guidance target person, for example, a list of personal IDs.
- the processor 25 includes a microprocessor such as a CPU and its peripheral circuits, and reads and executes the program 24P from the storage unit 24, thereby realizing the various processing units in cooperation with the hardware and the program 24P. have.
- main processing units realized by the processor 25 there are an input unit 25A, a discriminant model learning unit 25B, a condition creating unit 25C, and a health guidance target person selecting unit 25D.
- the input unit 25A, the discrimination model learning unit 25B, and the condition creation unit 25C have the same functions as the input unit 15A, the discrimination model learning unit 15B, and the condition creation unit 15C in the first embodiment.
- the health guidance target person selecting unit 25D reads the selection condition 24E and the health check data 24F for selection from the storage unit 24, and selects individuals who have health check data suitable for the selection condition 24E from the health check data 24F. And has a function of storing in the storage unit 24 as the selector 24G. In addition, the health guidance target person selecting unit 25D has a function of reading the selecting person 24G from the storage unit 24 and outputting the selected person 24G to the screen display unit 23 or outputting it to the outside through the communication I / F unit 21.
- the input unit 25A inputs the health check data 24A, the flag 24B, the insurer's desired condition 24C, the discrimination model 24D before learning, and the health check data 24F for selection from the communication I / F unit 21 or the operation input unit 22. And stored in the storage unit 24 (step S11).
- the discrimination model learning unit 25B reads the health diagnosis data 24A, the flag 24B, the insurer's desired condition 24C, and the discrimination model 24D before learning from the storage unit 24, and the health diagnosis data 24A, the flag 24B, and the insurer.
- the discriminant model 24D is learned using the desired condition 24C in the same manner as the discriminant model learning unit 15B in the first embodiment (step S12).
- the learned discrimination model 24 ⁇ / b> D is stored in the storage unit 24.
- condition creating unit 25C reads the learned discrimination model 24D from the storage unit 24, and among the plurality of explanatory variables of the learned discrimination model 24D, similarly to the condition creating unit 15C in the first embodiment, A combination of the health check item whose coefficient value is other than 0 and the coefficient value, and a determination threshold value are created as the selection condition 24E, stored in the storage unit 24, and output to the screen display unit 23. Alternatively, it is output to the outside through the communication I / F unit 21 (step S13).
- the health guidance target person selecting unit 25D reads the selection condition 24E and the health checkup data 24F for selection from the storage unit 24, and the health check data 24F from the health checkup data 24F is used to treat individuals who have health checkup data that meets the selection condition 24E.
- the person to be instructed is determined, stored in the storage unit 24 as the selector 24G, and output to the screen display unit 23 or output to the outside through the communication I / F unit 21 (step S14).
- the same effect as that of the first embodiment can be obtained, and the health guidance target person can be selected from the viewpoint of whether or not there is a possibility that the health guidance standard will be met in the next year. It becomes possible.
- the selection condition for the health guidance target person that meets the insurer's desired condition is generated.
- the selection condition may be generated without considering the insurer's desired condition.
- an objective function in which the third term on the right side in the above four formulas is omitted may be used.
- the number of health check items as selection conditions is reduced as much as possible.
- the determination threshold value is calculated, and an individual having a score equal to or higher than the determination threshold value is selected as the health guidance target person.
- the health guidance target person is used without using the judgment threshold value. May be selected. For example, it is possible to calculate the probability corresponding to the health guidance target person in the next fiscal year for each individual using the above formula 1, and select the top N persons with the probability as the health guidance target person.
Abstract
Description
第1の期間における個人の健康診断データである第1の健康診断データと、前記個人が前記第1の期間の次の期間である第2の期間の健康診断データで予め定められた保健指導基準に該当したか否かを表すラベル値とを記憶するメモリと、前記メモリに接続されたプロセッサとを備え、
前記プロセッサは、
前記健康診断データの複数の健康診断項目を複数の説明変数とし、前記説明変数と前記説明変数毎の係数とから構成される多項式で表現され、前記個人が前記第2の期間の健康診断データで前記保健指導基準に該当するか否かを判別するための判別モデルを、前記第1の健康診断データと前記ラベル値とを用いて学習し、
前記学習後の前記判別モデルにおける前記複数の説明変数としての前記複数の健康診断項目と前記係数の値との組み合わせを保健指導対象者選定条件として生成する
ようにプログラムされている、といった構成を採る。
また本発明の第2の観点にかかる保健指導対象者選定条件作成支援方法は、
第1の期間における個人の健康診断データである第1の健康診断データと、前記個人が前記第1の期間の次の期間である第2の期間の健康診断データで予め定められた保健指導基準に該当したか否かを表すラベル値とを記憶するメモリと、前記メモリに接続されたプロセッサとを備えた装置が実行する保健指導対象者選定条件作成支援方法であって、
前記プロセッサが、
前記健康診断データの複数の健康診断項目を複数の説明変数とし、前記説明変数と前記説明変数毎の係数とから構成される多項式で表現され、前記個人が前記第2の期間の健康診断データで前記保健指導基準に該当するか否かを判別するための判別モデルを、前記第1の健康診断データと前記ラベル値とを用いて学習し、
前記学習後の前記判別モデルにおける前記複数の説明変数としての前記複数の健康診断項目と前記係数の値との組み合わせを保健指導対象者選定条件として生成する、
といった構成を採る。
[第1の実施形態]
図1を参照すると、本発明の第1の実施形態にかかる保健指導対象者選定条件作成支援装置1は、保健指導基準に該当する見込みがあるかないかの観点から保健指導対象者を選定する条件を作成する機能を有している。
入力部15Aは、健康診断データ14Aと、フラグ14Bと、保険者の希望条件14Cと、学習前の判別モデル14Dとを入力する。
判別モデル14Dで求める値は、以下のPである。判別モデル学習部15Bは、この値を算出するための判別モデル14Dを用いて、そのパラメータを学習する。
P(Y=1|X;W)=1/(1+exp(W^{T}X)) …(1)
P(Y=0|X;W)=1-P(Y=1|X;W) …(2)
ただし、P(●|○;★)は★をパラメータとし、○が与えられた場合の●の条件付確率を表す。また、上付きのTはベクトルの転置を表す。
L(W)=\sum^{N}_{n=1}logP(Y_n|X_n,W) …(3)
ここで、\sum^{N}_{n=1}は、n=1からNまでの総和を表す。
ここで、||W||はWのノルムであり、ノルム1を用いる。Fは、N次元ベクトルで、n番目の要素はW^{T}X_nである。またL’は、正規化グラフラプラシアンである。λとαは、右辺の第1項と第2項と第3項のバランスを調整するパラメータである。
-(α/2){(W^{T}X_1 - W^{T}X_2)^2 + (W^{T}X_3)^2} …(5)
上記4式の目的関数を最大化するためには、保険者の希望する条件を満たさない個人3に係る(W^{T}X_3)^2の値が小さくなるようにW*を学習する必要がある。また、保険者の希望する条件を満足する個人1,2については、それら個人間でW^{T}Xの値が等しくなるようにW*を学習する必要がある。
図7を参照すると、本発明の第2の実施形態にかかる保健指導対象者選定条件作成支援装置2は、保健指導基準に該当する見込みがあるかないかの観点から保健指導対象者を選定する条件を作成する機能と、作成された選定条件に従って保健指導対象者を選定する機能とを有している。
以上、本発明を幾つかの実施形態を挙げて説明したが、本発明は以上の実施形態のみに限定されず、その他各種の付加変更が可能である。例えば、以下のような実施形態も本発明に含まれる。
11、21…通信I/F部
12、22…操作入力部
13、23…画面表示部
14、24…記憶部
15、25…プロセッサ
Claims (9)
- 第1の期間における個人の健康診断データである第1の健康診断データと、前記個人が前記第1の期間の次の期間である第2の期間の健康診断データで予め定められた保健指導基準に該当したか否かを表すラベル値とを記憶するメモリと、前記メモリに接続されたプロセッサとを備え、
前記プロセッサは、
前記健康診断データの複数の健康診断項目を複数の説明変数とし、前記説明変数と前記説明変数毎の係数とから構成される多項式で表現され、前記個人が前記第2の期間の健康診断データで前記保健指導基準に該当するか否かを判別するための判別モデルを、前記第1の健康診断データと前記ラベル値とを用いて学習し、
前記学習後の前記判別モデルにおける前記複数の説明変数としての前記複数の健康診断項目と前記係数の値との組み合わせを保健指導対象者選定条件として生成する
ようにプログラムされている保健指導対象者選定条件作成支援装置。 - 前記プロセッサは、
前記判別モデルの学習では、前記判別モデルの尤度を表す項と、値が0でない前記係数の個数に依存する罰則項とを有する目的関数を最適化するように前記判別モデルの前記係数の値を学習する
請求項1に記載の保健指導対象者選定条件作成支援装置。 - 前記メモリは、さらに、保険者の希望する条件を記憶し、
前記プロセッサは、
前記判別モデルの学習では、前記判別モデルの尤度を表す項と、値が0でない前記係数の個数に依存する罰則項と、前記保険者の希望する条件を満たさない前記個人が前記保健指導基準に該当することに依存する罰則項とを有する目的関数を最適化するように前記判別モデルの前記係数の値を学習する
請求項1に記載の保健指導対象者選定条件作成支援装置。 - 前記プロセッサは、
前記保健指導対象者選定条件の生成では、前記学習後の判別モデルにおける前記複数の係数のうち、値が0でない係数と当該係数に対応する説明変数としての前記健康診断項目との組み合わせを、前記保健指導対象者選定条件として生成する
請求項2または3に記載の保健指導対象者選定条件作成支援装置。 - 前記プロセッサは、
前記保健指導対象者選定条件の生成では、前記学習後の判別モデルにおける前記複数の係数のうち、値が0でない係数と当該係数に対応する説明変数としての前記健康診断項目との組み合わせと、前記学習後の判別モデルにおいて前記個人が前記第2の期間の健康診断データで前記保健指導基準に該当する確率が所定値以上であると判定できる、前記組み合わせに含まれる前記係数の値の合計値の最小値である判定しきい値とを、前記保健指導対象者選定条件として生成する
請求項2または3に記載の保健指導対象者選定条件作成支援装置。 - 前記メモリは、さらに、保健指導対象者候補である個人の健康診断データである第2の健康診断データを記憶し、
前記プロセッサは、さらに、
前記第2の健康診断データから、前記保健指導対象者選定条件に適合する前記個人を決定する
請求項1乃至5に記載の保健指導対象者選定条件作成支援装置。 - 前記メモリは、さらに、保健指導対象者候補である個人の健康診断データである第2の健康診断データを記憶し、
前記プロセッサは、さらに、
前記第2の健康診断データから、前記保健指導対象者選定条件に適合する前記個人を決定し、前記保健指導対象者選定条件に適合する前記個人の決定では、前記個人の前記第2の健康診断データ毎に、前記保健指導対象者選定条件中の前記健康診断項目のうちの該当する項目に対するスコアの総和を計算して前記判定しきい値と比較する
請求項5に記載の保健指導対象者選定条件作成支援装置。 - 第1の期間における個人の健康診断データである第1の健康診断データと、前記個人が前記第1の期間の次の期間である第2の期間の健康診断データで予め定められた保健指導基準に該当したか否かを表すラベル値とを記憶するメモリと、前記メモリに接続されたプロセッサとを備えた装置が実行する保健指導対象者選定条件作成支援方法であって、
前記プロセッサが、
前記健康診断データの複数の健康診断項目を複数の説明変数とし、前記説明変数と前記説明変数毎の係数とから構成される多項式で表現され、前記個人が前記第2の期間の健康診断データで前記保健指導基準に該当するか否かを判別するための判別モデルを、前記第1の健康診断データと前記ラベル値とを用いて学習し、
前記学習後の前記判別モデルにおける前記複数の説明変数としての前記複数の健康診断項目と前記係数の値との組み合わせを保健指導対象者選定条件として生成する
保健指導対象者選定条件作成支援方法。 - 第1の期間における個人の健康診断データである第1の健康診断データと、前記個人が前記第1の期間の次の期間である第2の期間の健康診断データで予め定められた保健指導基準に該当したか否かを表すラベル値とを記憶するメモリに接続されたプロセッサに、
前記健康診断データの複数の健康診断項目を複数の説明変数とし、前記説明変数と前記説明変数毎の係数とから構成される多項式で表現され、前記個人が前記第2の期間の健康診断データで前記保健指導基準に該当するか否かを判別するための判別モデルを、前記第1の健康診断データと前記ラベル値とを用いて学習するステップと、
前記学習後の前記判別モデルにおける前記複数の説明変数としての前記複数の健康診断項目と前記係数の値との組み合わせを保健指導対象者選定条件として生成するステップと、
を行わせるためのプログラム。
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US14/002,550 US20150074019A1 (en) | 2012-04-26 | 2013-04-03 | Health guidance receiver selection condition generation support device |
SG11201406704RA SG11201406704RA (en) | 2012-04-26 | 2013-04-03 | Health guidance receiver selection condition generation support device |
JP2013534517A JP5392444B1 (ja) | 2012-04-26 | 2013-04-03 | 保健指導対象者選定条件作成支援装置 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2012100937 | 2012-04-26 | ||
JP2012-100937 | 2012-04-26 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2013161191A1 true WO2013161191A1 (ja) | 2013-10-31 |
Family
ID=49482546
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2013/002301 WO2013161191A1 (ja) | 2012-04-26 | 2013-04-03 | 保健指導対象者選定条件作成支援装置 |
Country Status (4)
Country | Link |
---|---|
US (1) | US20150074019A1 (ja) |
JP (1) | JP5392444B1 (ja) |
SG (1) | SG11201406704RA (ja) |
WO (1) | WO2013161191A1 (ja) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105808906A (zh) * | 2014-10-27 | 2016-07-27 | 三星Sds株式会社 | 患者个人特性的分析方法及其装置 |
JP2016218966A (ja) * | 2015-05-26 | 2016-12-22 | 株式会社日立製作所 | 分析システム、及び、分析方法 |
JP2017117469A (ja) * | 2015-12-22 | 2017-06-29 | 国立研究開発法人理化学研究所 | リスク評価方法、リスク評価装置及びリスク評価プログラム |
JP2020071562A (ja) * | 2018-10-30 | 2020-05-07 | 株式会社キャンサースキャン | 健康診断受診確率計算方法及び健診勧奨通知支援システム |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR102037573B1 (ko) * | 2017-03-09 | 2019-10-29 | 주식회사 파트너스앤코 | 문진 데이터 및 카메라 데이터에 기반한 진단용 데이터 처리 장치 및 그 시스템 |
JP7140410B2 (ja) * | 2018-03-30 | 2022-09-21 | Necソリューションイノベータ株式会社 | 予測システム、予測方法および予測プログラム |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2006301893A (ja) * | 2005-04-20 | 2006-11-02 | Hitachi Ltd | 保健事業支援システム |
JP2007257565A (ja) * | 2006-03-27 | 2007-10-04 | Hitachi Ltd | 保健事業支援システム |
JP2009151404A (ja) * | 2007-12-19 | 2009-07-09 | Hitachi Ltd | 保健指導対象者選定支援システム |
JP2009151423A (ja) * | 2007-12-19 | 2009-07-09 | Katsuzo Kawanishi | 保健指導支援装置 |
JP2011065579A (ja) * | 2009-09-18 | 2011-03-31 | Nec Corp | 標準パタン学習装置、ラベル付与基準算出装置、標準パタン学習方法およびプログラム |
Family Cites Families (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2002063279A (ja) * | 2000-08-23 | 2002-02-28 | Nec Corp | 健康管理のためのシステム及び方法並びに健康管理用プログラムを記憶した記憶媒体 |
JP4062910B2 (ja) * | 2001-11-29 | 2008-03-19 | 株式会社日立製作所 | 健康管理支援方法及び装置と健康余命予測データ生成方法及び装置 |
JP4470679B2 (ja) * | 2004-10-07 | 2010-06-02 | 株式会社日立製作所 | 健康指導支援システム |
US20070259377A1 (en) * | 2005-10-11 | 2007-11-08 | Mickey Urdea | Diabetes-associated markers and methods of use thereof |
AU2007308078A1 (en) * | 2006-10-13 | 2008-04-17 | PeraHealth, Inc | System and method for providing a health score for a patient |
WO2009001862A1 (ja) * | 2007-06-25 | 2008-12-31 | Ajinomoto Co., Inc. | 内臓脂肪蓄積の評価方法 |
WO2010048282A1 (en) * | 2008-10-21 | 2010-04-29 | Rothman Healthcare Research, Llc | Methods of assessing risk based on medical data and uses thereof |
US8346688B2 (en) * | 2009-11-25 | 2013-01-01 | International Business Machines Corporation | Predicting states of subjects |
-
2013
- 2013-04-03 WO PCT/JP2013/002301 patent/WO2013161191A1/ja active Application Filing
- 2013-04-03 JP JP2013534517A patent/JP5392444B1/ja active Active
- 2013-04-03 US US14/002,550 patent/US20150074019A1/en not_active Abandoned
- 2013-04-03 SG SG11201406704RA patent/SG11201406704RA/en unknown
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2006301893A (ja) * | 2005-04-20 | 2006-11-02 | Hitachi Ltd | 保健事業支援システム |
JP2007257565A (ja) * | 2006-03-27 | 2007-10-04 | Hitachi Ltd | 保健事業支援システム |
JP2009151404A (ja) * | 2007-12-19 | 2009-07-09 | Hitachi Ltd | 保健指導対象者選定支援システム |
JP2009151423A (ja) * | 2007-12-19 | 2009-07-09 | Katsuzo Kawanishi | 保健指導支援装置 |
JP2011065579A (ja) * | 2009-09-18 | 2011-03-31 | Nec Corp | 標準パタン学習装置、ラベル付与基準算出装置、標準パタン学習方法およびプログラム |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105808906A (zh) * | 2014-10-27 | 2016-07-27 | 三星Sds株式会社 | 患者个人特性的分析方法及其装置 |
JP2016218966A (ja) * | 2015-05-26 | 2016-12-22 | 株式会社日立製作所 | 分析システム、及び、分析方法 |
JP2017117469A (ja) * | 2015-12-22 | 2017-06-29 | 国立研究開発法人理化学研究所 | リスク評価方法、リスク評価装置及びリスク評価プログラム |
JP2020071562A (ja) * | 2018-10-30 | 2020-05-07 | 株式会社キャンサースキャン | 健康診断受診確率計算方法及び健診勧奨通知支援システム |
Also Published As
Publication number | Publication date |
---|---|
JPWO2013161191A1 (ja) | 2015-12-21 |
SG11201406704RA (en) | 2014-11-27 |
US20150074019A1 (en) | 2015-03-12 |
JP5392444B1 (ja) | 2014-01-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Cutler et al. | Physician beliefs and patient preferences: a new look at regional variation in health care spending | |
Swaminathan et al. | A machine learning approach to triaging patients with chronic obstructive pulmonary disease | |
KR101855117B1 (ko) | 질병에 대한 발병 확률 예측 방법 및 장치 | |
JP5392444B1 (ja) | 保健指導対象者選定条件作成支援装置 | |
Chao et al. | Principal component-based weighted indices and a framework to evaluate indices: Results from the Medical Expenditure Panel Survey 1996 to 2011 | |
Laacke et al. | Artificial intelligence, social media and depression. A new concept of health-related digital autonomy | |
Hamilton et al. | Designed to fit: The development and validation of the STRONG-R recidivism risk assessment | |
Robinson et al. | Participants and study decliners’ perspectives about the risks of participating in a clinical trial of whole genome sequencing | |
KR101885111B1 (ko) | 질환 발병 예측 방법 및 장치 | |
Wilson et al. | Literacy, cognitive ability, and the retention of health-related information about colorectal cancer screening | |
Dalakleidi et al. | Comparative assessment of statistical and machine learning techniques towards estimating the risk of developing type 2 diabetes and cardiovascular complications | |
Campbell et al. | Statistical innovations in the medical device world sparked by the FDA | |
Klest et al. | Trauma, socioeconomic resources, and self-rated health in an ethnically diverse adult cohort | |
Barbeira et al. | Widespread dose-dependent effects of RNA expression and splicing on complex diseases and traits | |
Chan et al. | Harmonizing measures of cognitive performance across international surveys of aging using item response theory | |
Choudhury et al. | Evaluating patient readmission risk: a predictive analytics approach | |
Davis et al. | Addressing geographic confounding through spatial propensity scores: a study of racial disparities in diabetes | |
Chapman et al. | Health risk prediction models incorporating personality data: Motivation, challenges, and illustration. | |
Roberts et al. | Genetic duties | |
Whiteman et al. | Transforming mental health services to address gender disparities in depression risk factors | |
Salmina | Social attitudes towards welfare policies in Russia and other European countries | |
Pirracchio et al. | Collaborative targeted maximum likelihood estimation for variable importance measure: Illustration for functional outcome prediction in mild traumatic brain injuries | |
Kundi et al. | Artificial Intelligence and Bias: A scoping review | |
Hans et al. | Boosting distributional copula regression | |
Usmani et al. | Artificial Intelligence Applications in Healthcare |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
ENP | Entry into the national phase |
Ref document number: 2013534517 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 14002550 Country of ref document: US |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 13781005 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 13781005 Country of ref document: EP Kind code of ref document: A1 |