JP2006113991A - Extraction method for personal health care information - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an extraction method for personal health care information useful for system users to perform health care based on health information in daily life time-sequentially stored in a personal health care system. <P>SOLUTION: Time-sequential personal health information consists of one or more health data items acquired through measurement of the body of a system user and one or more life data items that indicate life information on the user. Based on a time-sequential correlation between health data measured at an arbitrary date or changes or the rate of change of health data for an arbitrary period, and an addition value or an average value of life data for an arbitrary period or its derivation value, correlations between the health data items and the life data items are extracted. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

Detailed Description of the Invention

  The present invention relates to a method for extracting personal health management information in a personal health management system that performs health management based on daily health information of users.

  Most of the conventional health care is performed based on health information obtained through regular health checkups and a medical checkup. However, these pieces of health information are data that are cut out only at the time when a regular health checkup and a medical checkup are performed. Therefore, there is a limit in extracting health management information for daily health management from only these data. Therefore, a system has been devised for the purpose of accumulating health information in the daily living environment and using it for primary prevention of diseases and health promotion.

  For example, in Patent Document 1, daily health information measured by a measuring device attached to a user is managed by a health management information center, and the requested health information is provided to the user to provide the user's health. Health information management methods that support management have been proposed.

  JP 2002-245177 A (Claim 1, FIG. 1)

Problems to be solved by the invention

  However, although the invention of Patent Document 1 describes the configuration and function of a system for managing health information, a method related to extraction of personal health management information that is useful for the user of the system to perform health management. Is not mentioned.

  Accordingly, an object of the present invention is to provide a method for extracting personal health management information that is useful for a user to perform health management based on daily health information accumulated in time series.

Means for solving the problem

  The method for extracting personal health management information according to the present invention, which has been made to solve the above-mentioned problems, includes one or more health data items representing biological information obtained by measuring the body of a user of the system, and the system In the time series personal health information consisting of one or more life data items representing the life information of users, the health data measured at any date or the change or rate of change of health data in any period, and any It is characterized in that a correlation between health data items and life data items is extracted by looking at a time series correlation with an addition value or an average value of life data over a period or a value derived therefrom or a pattern of life data.

ここで、Δｐ_ｎｍはある任意の期間における健康データの変化である。これを任意のある期間で割れば健康データの変化率になる。ｅ^ｔ _ｉｊはある任意の期間における生活データの加算（単

ｎ日とｉ日は一般的に同日ではなく、図１に示したようにオフセット期間ｓを定義する。また任意の期間ｎ−ｍとｉ−ｊは必ずしも同一期間である必要はない。次に、任意の期間ｎ−ｍとｉ−ｊ，およびオフセット期間ｓをパラメータとし、多くの時系列データを基にｎ，ｉを変化させ

を採用する。このとき相関係数は、任意の期間ｎ−ｍとｉ−ｊ，オフセット期間ｓの値によって変化する。例えば最大の相関を示すｎ−ｍ，ｉ−ｊ，ｓの組み合わせを（ｎ−ｍ）_ｍａｘ，（ｉ−ｊ）_ｍａｘ，ｓ_ｍａｘとすると、（ｎ−ｍ）_ｍａｘと（ｉ−ｊ）_ｍａｘが大きいほど、生活データ項目の長期間にわたるデータが健康データ項目の現在値に大きな影響を与えることになり、小さいほど生活データ項目の短期間におけるデータが健康データ項目の現在値に大きな影響を与えることになる。また、ｓ_ｍａｘが大きほど生活データ項目の影響が遅れて健康データ項目のデータに現れ、小さいほど生活データ項目の影響がすぐに健康データ項目のデータに反映されるということになる。すなわち、このような時系列データの解析により生活データ項目と健康データ項目の相関関係をみることができ、個人の健康管理に有用な個人健康管理情報を抽出することができる。FIG. 1 is a diagram showing the concept of the personal health management information extraction method of the present invention. _pn is the data on day _n of a certain health data item, and ei is the data on day _i of a certain life data item. Each n days, the few days before the data i date and p _m, e _j, for example, define the following quantities.
_{Δp nm = p n -p m (} 1)
e ^t _ij = e _i + e _i−1 +... + e _j (2)

Here, Δp _nm is a change in health data in an arbitrary period. Dividing this by any given period gives the rate of change in health data. e ^t _ij is an addition of life data in a certain arbitrary period (simple

The days n and i are generally not the same day, but an offset period s is defined as shown in FIG. Further, the arbitrary periods nm and ij are not necessarily the same period. Next, with arbitrary periods nm and ij and offset period s as parameters, n and i are changed based on a large amount of time-series data.

Is adopted. At this time, the correlation coefficient changes depending on the values of arbitrary periods nm and ij and the offset period s. For example, if the combination of nm, ij, and s showing the maximum correlation is (nm) _max , (ij) _max , _smax , (nm) _max and (ij) _max The longer the life data item, the longer the life data item has a greater effect on the current value of the health data item, and the smaller the life data item, the shorter the life data item, the greater the current value of the health data item. It will be. In addition, the larger the s _max , the later the influence of the life data item appears in the data of the health data item, and the smaller the s _max , the sooner the influence of the life data item is reflected in the data of the health data item. That is, by analyzing such time series data, the correlation between life data items and health data items can be seen, and personal health management information useful for personal health management can be extracted.

Incidentally, Healthy not necessarily change or rate of change in any period for data, values or lifestyle data is derived sum of life data or an average value or therefrom in the measured health data p _n and an arbitrary period in a certain time It is also effective to look at the time series correlation with the pattern. It is also effective to use weighted addition values and average values instead of simple additions and averages as addition values and average values of life data in an arbitrary period. Furthermore, it is also effective to use values obtained by deriving from the added value or average value of life data items, for example, the reciprocal of these values, or a combination of values for a plurality of life data items.

  Hereinafter, embodiments of the present invention, blood pressure and body fat percentage as health data items, calorie consumption due to exercise, calorie intake due to diet, alcohol intake (qualitative value), stress (qualitative value), and health data items, The sleep time will be described in detail as an example. Data for these items is recorded daily for approximately three months by users of certain systems for health care.

In the first example, we are interested in how life data items such as calories burned by exercise, alcohol intake, stress, and sleep time affect daily blood pressure, and the correlation between each life data item and blood pressure Was analyzed by the method described above. FIG. 2 shows values of (i−j) _max and s _max that show the maximum correlation with time series data of minimum blood pressure values for each life data item. Hereinafter, FIG. 2 will be described.

(1) Calories consumed by exercise:
The simple addition value of the data on the day before and 2 days before the blood pressure measurement day showed the maximum correlation with the minimum blood pressure value. This added value is indicated as total calorie consumption 2.

(2) Alcohol intake:
The added value of the data 2 days before and 3 days before the blood pressure measurement day showed the maximum correlation with the minimum blood pressure value. In this example, since the alcohol intake is expressed in five levels of 1 to 5 (1 is the largest and 5 is the smallest), the reciprocal of the number representing the degree is used as data. For example, when 2 days are 3 and 4, the alcohol intake is 1/3 + 1/4 = 0.58. This added value is indicated as alcohol intake 2.

(3) Stress:
The reciprocal of the average value of the data from the previous day to the third day before the blood pressure measurement day showed the maximum correlation with the minimum blood pressure value. In this example, the stress is expressed in three grades of 1 to 3 (1 high, 2. normal, 3. low), so the reciprocal of the average value was taken. For example, when 3 days is (2. normal, 3. less, 3. less), stress = 1 / ((2 + 3 + 3) / 3) = 0.38. This value is indicated as stress 3.

ている。(4) Sleep time:
The weighted addition value of the data from the day before the blood pressure measurement day to 5 days before showed the maximum correlation with the minimum blood pressure value. The weight here is due to the depth of sleep. In this example, the depth of sleep is expressed in three stages of 1 to 3 (1. Good sound, 2. Slightly shallow, 3. I couldn't sleep well). I put on. Specifically, the sleep time of the day was divided by the square root of a number indicating the degree of sleep depth. For example, sleep time is 7 hours, but sleep is a little

ing.

  Next, the added value or average value of the above-mentioned life data in the period showing the maximum correlation with the minimum blood pressure value shown in FIG. 2 or a derived value thereof is used as an input variable, and the minimum blood pressure itself is used as a target variable for about 3 months. A decision tree generation algorithm is applied to the data to generate a decision tree. The obtained decision tree is shown in FIG. Here, in order to use the minimum blood pressure, which is numerical data, as a target variable of the decision tree, the data is divided into three categories of “high”, “intermediate”, and “low” depending on the size of the minimum blood pressure. The root record is divided into two nodes according to the total calorie consumption 2, and the node classified into the larger one is divided into two leaves according to the magnitude of the stress 3. One node is further divided into nodes and leaves according to alcohol intake 2 and sleep time 5.

The correlation rule between the minimum blood pressure extracted from this decision tree and the above life data item is expressed as follows.
(1) If the sleeping time 5 is 25.1 hours or less and the total calorie consumption 2 is 738 kcal or less, the minimum blood pressure tends to increase.
(2) If the sleeping time 5 is larger than 25.1 hours and the total calorie consumption 2 is 738 kcal or less, the minimum blood pressure tends to be intermediate.
(3) When the stress 3 is 0.38 or less and the total calorie consumption 2 is larger than 738 kcal, the minimum blood pressure tends to be low.
(4) If the alcohol intake 2 is 0.58 or less, the minimum blood pressure tends to be low.
These are personal health management information useful for daily health management of individual users of the system, and based on this, the user of the system can improve lifestyle with a goal.

The next example is how life data items such as calories consumed by exercise and calories consumed from meals affect daily body fat percentage, and the correlation between each life data item and body fat percentage Was analyzed by the method described above. The values of (i−j) _max and s _max that show the maximum correlation with the time series data of the body fat percentage for each life data item are shown in FIG. Hereinafter, FIG. 4 will be described.

(1) Calories consumed by exercise:
The simple addition value of the data from the day before the body fat percentage measurement day to 5 days before showed the maximum correlation with the body fat percentage. This added value is indicated as total calorie consumption 5.

(2) Dietary calorie intake:
The calorie intake by the meal 2 days before the body fat percentage measurement day showed the maximum correlation with the change in the body fat percentage from 2 days before. Here, the calorie intake from the meal is the total value of the calorie intake in the morning, noon and night of the day. This value is displayed as 1 total calorie intake.

From this example regarding body fat percentage,
(1) The body fat percentage tends to decrease by increasing the total total calories burned for 5 days. In other words, even if you exercise for one day or two days, your body fat percentage will not decrease immediately.
(2) Excessive daily food intake tends to appear as an increase in body fat percentage after 2 days.
These are personal health management information useful for individual health management of the user of the system, and based on this, the system user can improve the lifestyle with a goal.

  The method for extracting personal health management information according to the present embodiment described above analyzes the time series data by comparing the health data item of interest with one or more life data items in advance, and the correlation between the health data item and the life data item. Is heading. That is, a decision tree is generated with the life data item as an input variable and the health data item as a target variable, and a correlation rule between the life data item and the health data item is extracted therefrom. However, a similar correlation rule is that data at a certain time of one or more health data items or a change or rate of change of data in an arbitrary period and an added value of data in an arbitrary period of one or more life data items or It can be easily inferred that it can also be obtained by performing time-series association rule analysis with a data set consisting of an average value, a value derived therefrom, or a data pattern as one transaction. Here, the pattern of the life data in an arbitrary period is added as a candidate for the data member in the data set, and the data is expressed in a qualitative degree like the alcohol intake and stress in the present embodiment. In some cases, it is possible to perform correlation rule analysis by recognizing data for three days as a pattern such as (1, 3, 2) or (2, 1, 1) without converting to numerical values. is there.

  In addition, the input variable of the decision tree composed of the life data items in the personal health management information extraction method of the present embodiment includes the addition value or average of the life data in the period showing the maximum correlation with the target variable composed of the health data items, respectively. A value or a value derived from it is used, but the input variable for generating the decision tree is not limited to this. This is because there may be a set of input variables defined by surrounding periods showing the maximum correlation for classifying records more efficiently, even if each is not a period showing the maximum correlation with the target variable consisting of health data items.

  Furthermore, the personal health management information extraction method of the present embodiment takes blood pressure and body fat percentage as health data items, calories consumed by exercise, calories consumed by diet, alcohol intake, stress, sleep time as examples of life data items. However, it is obvious that other items can be applied. For example, in the health data item, it is clear that the smoking amount and the like can be targeted in the life data item such as body weight, body temperature, pulse rate, blood sugar level.

The invention's effect

  According to the present invention, personal health management information useful for personal health management can be extracted from personal health data and daily life data accumulated in time series in daily life. Can improve lifestyle habits for primary prevention and health promotion.

    It is a conceptual diagram of the extraction method of the personal health management information of this invention. It is a figure which shows the period (ij) _{max of the} life data which shows the maximum correlation with the minimum blood pressure value, and offset period _smax .     It is a figure which shows the example of the decision tree which used the minimum blood pressure as a target variable. From this decision tree, a correlation rule between the minimum blood pressure and the life data item can be extracted. It is a figure which shows the period (ij) _{max of} life data which shows the maximum correlation with a body fat rate, and offset period _smax .

Claims (7)

  A time-series individual consisting of one or more health data items representing biological information obtained by measuring the body of a person performing health care and one or more life data items representing life information of the person performing health care In health information, health data measured at any date or change or rate of change in health data in any period, and the sum or average value of life data in any period or values derived from it or patterns of life data A method for extracting personal health management information, characterized in that a correlation between health data items and life data items is extracted by looking at the time-series correlations.   2. The method for extracting personal health management information according to claim 1, wherein the addition value or average value of the life data is a weighted addition value or average value.   A data value measured at an arbitrary date and time of the health data item or data in an arbitrary period using an addition value or an average value of data in an arbitrary period of one or more life data items or a value derived therefrom as an input variable 3. The method of extracting personal health management information according to claim 1, wherein a decision tree is generated using a change or rate of change as a target variable.   4. The method for extracting personal health management information according to claim 3, wherein a correlation rule between the health data item and the life data item is extracted based on the decision tree.   An added value or average value of data in an arbitrary period of one or more life data items, or a value or data pattern derived therefrom, and a data value measured at an arbitrary date and time of one or more health data items, or 2. A time series correlation rule analysis is performed by using a data set composed of a change or rate of change of data in an arbitrary period as one time series transaction, and a correlation rule having a premise part and a conclusion part is extracted. The method for extracting personal health management information according to claim 2.   6. The method of extracting personal health management information according to claim 1, wherein the health data item is at least one of body weight, body fat percentage, body temperature, blood pressure, pulse rate, and blood glucose level.   6. The life data item is at least one of calorie consumption by exercise, calorie intake by meal, alcohol intake, smoking amount, sleep time, sleep depth, and stress. The method of extracting the personal health management information described.

Images