JP2007328751A - Data analysis processing device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a data analysis processing device rearranging time axes of samples in advance on the basis of time zones with event change, and efficiently carrying out analysis processing using the same time axis. <P>SOLUTION: In statistical analysis of events over time, the data analysis processing device is characterized by that it has a mechanism calculating time zones of samples with event change over time for representing existence of events in time zones divided on the time axis, and analyzing events of different time zones with respect to a plurality of samples. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a data analysis processing apparatus, a database, an information processing method, and a program recording medium for efficiently performing statistical analysis of events over time.

  When statistically analyzing events over time, it is necessary to re-analyze the time axis of each sample in advance based on the time interval in which the event change occurred over time, and then perform the analysis process. was there. Since there has been no data analysis processing apparatus that efficiently performs such preprocessing, analysis of a large amount of data has been difficult because various event changes are analyzed particularly in interactive online analysis with a user.

  In addition, due to the nature of sample surveys, statistical significance can be verified and samples can be used effectively even if there are missing or incomplete data in some time intervals, but additional calculation processing is required for that purpose. was there.

  For example, the prescription data and test data of the same patient are analyzed over time for the purpose of investigating the usefulness and side effects of drugs, but there are prescriptions from multiple doctors and medical institutions. Depending on the patient's condition, there may be advance medical treatment, prescription addition, or change. Due to such complexity, data inadequacies, and deficiencies, the task of aligning time intervals where changes have been made to prescriptions or changes in test values for each patient is currently being done manually. This is a problem for efficient analysis.

The present invention is a data processing control apparatus that solves a problem that has not been solved by the conventional technology described above.
The problem with the conventional technology is that, when statistically analyzing events over time, the time axis of each sample is aligned in advance based on the time interval in which the event change occurred, and the analysis process using the same time axis is efficient. Since there is no automatic data analysis processing device, processing and work including verification of statistical significance of sample data is inefficient.

  In order to effectively solve the problems of the conventional technology, the present invention expresses statistical analysis of events over time as the presence or absence of an event in a segmented time interval on the time axis, and different time intervals for a plurality of samples A data analysis processing apparatus, database, information processing method, and program recording medium having a mechanism for calculating a time interval of each sample in which an event change with time has occurred.

  According to the present invention, when statistically analyzing events over time, analysis processing can be performed mechanically and efficiently, and information analysis can be performed statistically and effectively in a sample survey.

  In the best mode for carrying out the present invention, an apparatus configuration and an information processing method of the present invention will be described as an example. In the following examples, pharmaceutical prescriptions and test information may be described as examples, but the present invention is not limited to medical information and can be implemented.

FIG. 1 is a basic apparatus configuration of the present invention, and FIG. 2 is a flowchart showing a basic information processing method of the present invention.
The data analysis processing apparatus 102 of FIG. 1 constitutes a conventional data analysis processing apparatus together with the specimen database 103. Events and samples to be analyzed are recorded in the sample database 103 together with related master information. The temporal event analyzer 104 is an information processing mechanism that characterizes the present invention. The temporal event analyzer 104 operates the specimen database 103 to calculate an event change occurrence time interval for each specimen, and records and holds the result in the temporal event database 105. Or The data analysis processing device 102 directly operates the sample database 103 or instructs the time-dependent event analysis device 104 to obtain and output the analysis result 106.
The information processing method will be described below along the flowchart of FIG.

The analysis command receiving step 201 in FIG. 2 corresponds to the data analysis processing apparatus 102 receiving the analysis command 101 in FIG.
The data analysis processing device 102 sends an analysis command 101 to the time-dependent event analysis device 104.

  In step 202, the time-dependent event analysis apparatus 104 determines whether or not the time-period event database 105 records and holds information on the time interval in which the event change to be analyzed occurs based on the received analysis command. If it is held, the process proceeds to step 203. If not, the process proceeds to step 204.

  In step 203, the stored information is retrieved and acquired, and the process proceeds to step 207.

  In step 204, the time interval length is set for the event to be analyzed. Definition information related to the time interval length can be recorded in the program of the time event analyzer 104, in memory, or as master information of the time event database 105.

  For example, in the case of drugs, the time interval is sufficient for a week for drugs taken for several months or longer, and the time length is 2 days or 3 days for antibiotics.

  When the time axis is a weekly interval, the week including the prescription date as the interval, the week including the day with the number of prescription days added as the interval, and the week existing in between are calculated as the interval where the prescription existed effectively can do. In addition, in both end sections, it is possible to take a calculation method in which a prescription date of four days or more is valid and invalid if it is three days or less.

  In addition, when the definition about a time interval length is not defined beforehand, it is good also as a value which calculates | requires the minimum value of the duration of an event from a sample, or covers the fixed ratio of a sample. In the latter case, usually, for example, a range of 2 to 3 times the mean ± standard deviation is often set statistically.

  In addition, when more accurate analysis of the duration of the event is required, the section length is further reduced to one half to improve the reliability of the duration in the output result. Such definition information can also be recorded in the program of the time event analyzer 104, in memory, or as master information of the time event database 105.

  In order to simplify the following description, it is assumed that the time interval is a natural integer expression from the past to the future.

  In step 205, quantitative events are also distinguished as events so that the processing can be simplified and mechanically processed, and are re-expressed as presence / absence of events for each sample and for each time interval. Express and remember.

  For example, in the prescription of pharmaceuticals, for each patient, the combination of the name of the drug or active ingredient and the daily dose or single dose is an event. The aspirin daily dosage of 1000 mg and the aspirin daily dosage of 1500 mg are distinguished as different events. If the time interval is expressed in units of weeks, patient A, aspirin daily dose 1000 mg, etc. for the week including the prescription date and the week including the day with the prescription day plus the prescription days, etc. To express.

  By this simplification, event expressions may overlap in the same time interval. As a matter of fact, the prescription has a slight overlap with the previous prescription, and if it is expressed in weekly time, if there is a prescription change, both the pre-change and the post-change are expressed in duplicate. . In the previous example of aspirin, if the aspirin is changed from 1000 mg to 1500 mg aspirin, the week of the change can be expressed as both 1000 mg aspirin and 1500 mg aspirin, that is, 2500 mg aspirin. It can be removed in step 210 of outputting the analysis result.

  The temporal event database 105 and the specimen database 103 may be physically the same database, magnetic disk, or other storage device.

  In step 206, the time interval in which the event change of each sample occurs is calculated, and recorded and retained in the time-dependent event database 105.

  The time interval in which an event change with time of each sample occurs can be obtained by simple maximum and minimum value search of the time interval in which search conditions are specified for the event change and the incidental conditions.

For example, the minimum value of the time interval in which a drug was prescribed, i.e., the oldest time interval represents the time interval in which the drug was started, and the maximum value of the time interval in which a drug was prescribed, i.e., the latest time. The section represents a time section in which the prescription is stopped.
It is also possible to search for an event that a prescribed medicine has not been prescribed for a certain period of time and has been prescribed thereafter, using the standard database operation language SQL.

  Although the time interval in which an event change has actually occurred varies from sample to sample, the time-dependent event analysis device 104 performs the subsequent processing by rearranging the time intervals using the time interval in which the event change has occurred as a reference, for example, 0. Further, the time interval in which the event change occurs is recorded and held in the time-dependent event database 105 for each sample. For example, when an event change corresponding to 12 weeks in the time interval occurs in a certain sample A, the subsequent processing is performed on the time axis in which 12 weeks is set to 0 week, that is, 12 is subtracted from the time interval.

  In sample surveys, it may not be possible to obtain information for all time intervals for individual samples. For example, when creating a medical or epidemiological lifetime model, it is not easy to follow an individual for decades. In addition, in individual sample units, data in some time sections may be invalid or missing, and information on all time sections may not be obtained.

  However, even if the samples that make up each time interval are not constant, for a sample that meets a certain event change and incidental conditions, multiple cases are arranged with reference to the time interval in which the event change occurred, and various events appear. Analyze frequency statistically to understand statistically significant trends or exceed the limits of individual sample analysis.

ただし、
Ｎは母数（母集団の件数）
ｐは頻度（事象の出現頻度）
ｄは信頼度に応じた係数（信頼度９５％では１．９６となる）
ｅは推定誤差（推定誤差の許容値）
ｎは標本数（以上の条件における必要な標本数）Therefore, as necessary, it is possible and useful to verify whether the number of samples in each time interval has reached an effective number in the statistical analysis of the target event using the formula in statistics, Equation 1. The verification result is recorded in the time-dependent event database 105, and is output together with other results in step 210 for outputting the analysis result. If the number of samples has not reached the effective number, the processing is interrupted or the interruption is performed by the user. It is also possible to make inquiries. In addition to the verification of effectiveness, it is also possible to apply Equation 1 to obtain the estimation error and reliability.

However,
N is the population (number of populations)
p is frequency (frequency of occurrence of events)
d is a coefficient corresponding to the reliability (when the reliability is 95%, it is 1.96)
e is the estimation error (tolerance value of the estimation error)
n is the number of samples (necessary number of samples under the above conditions)

  From step 209 for calculating the appearance frequency of the event, the conventional data analysis device 102 and the sample based on the basis of the time interval for each sample related to the event change output from the time-dependent event analysis device 104 and the time-dependent event database 105. From the database 103, analysis results such as the appearance frequency of various events for each time interval are calculated. This method for calculating the appearance frequency is to obtain a value obtained by dividing the number of samples having an event by the total number of samples, and is the same as the conventional method.

  In step 210 for outputting the analysis result, the analysis result of step 209 for calculating the appearance frequency of the event is simply output. However, as described above, different events occur in the time interval in which the event change occurs. There is a possibility that expressions are duplicated. In order to eliminate the adverse effects, the following processing can be performed.

  Focusing on the fact that the overlap of events is resolved over time, the difference between the sections before and after the event change is output as an analysis result. If the changed week is defined as 0 week, the difference between -1 week, which is the immediately preceding week, and +1 week, which is the immediately following week, is output. Also, when outputting as time-lapse data, the zero-week data is skipped. Similarly, depending on the purpose and use, the results of the sections other than immediately before and after, for example, -2 weeks and +2 weeks, and the results of the preceding and following multiple sections, for example, the average of -2 weeks to -1 weeks and +1 week to +2 weeks May be output. The definition of the output of the analysis result can be recorded as master information of the temporal event database 105 in the program and memory of the temporal event analyzer 104.

The analysis processing of the present invention described with reference to the flowchart of FIG. 2 may be practically used as a common program stored in the specimen database 103 or the temporal event database 105, but the logical device configuration is as follows. It is not different from FIG.
The present invention can be implemented as a batch process or an online process. However, in the online process, the user looks at the analysis result, changes the analysis conditions little by little, and repeatedly performs the analysis process. The improvement effect is particularly high.

As described in the present embodiment, the time interval length is appropriately set, and the simplification of re-expressing the presence or absence of an event makes it possible to mechanically calculate the occurrence time interval of the event change over time, In addition, it is possible to output the analysis result by removing the adverse effects of simplification.
For example, the present invention efficiently analyzes and processes a daily dose of a drug, a relationship between a taking period and a test value, a correlation, information on a drug that is additionally prescribed after the start of prescription of a certain drug, and the like.

Basic apparatus configuration of the present invention Flow chart showing analysis process

Explanation of symbols

101 Analysis Command 102 Data Analysis Processing Device 103 Sample Database 104 Time-lapse Event Analysis Device 105 Time-lapse Event Database 106 Analysis Result 201 Analysis Command Acceptance Step 202 Step for Confirming Retention of Event Change Occurrence Time Section 203 Step for Retrieving Retention Information 204 Time Step of setting interval length 205 Step of expressing presence / absence of event in time interval 206 Step of calculating and holding event change occurrence time interval of each sample Step of aligning time interval of 207 samples Verifying statistical significance by 208 number 1 Step 209 Step of calculating the appearance frequency of the event 210 Step of outputting the analysis result

Claims (6)

  For statistical analysis of events over time, each occurrence of event change over time is represented to represent the presence or absence of events in segmented time intervals on the time axis and analyze events in different time intervals for multiple samples. A data analysis processing apparatus having a mechanism for calculating a time interval of a sample.   2. The data analysis processing apparatus according to claim 1, wherein the time interval of each sample in which a specific event change occurs is calculated and held in advance, and the analysis process is efficiently performed.   3. The data analysis processing apparatus according to claim 1, wherein an analysis result is output for an event in one or a plurality of preceding and following sections of a time section in which an event change with time occurs.   3. The data analysis processing apparatus according to claim 1, further comprising a mechanism for calculating statistical significance, estimation error, and error rate for the number of samples in each time interval.   3. The data analysis processing apparatus according to claim 1, wherein, for the whole sample or a part of the sample, the minimum value of the duration of the event or a half thereof is set as the length of the time interval.   The data according to claim 1 or 2, wherein a definition of a time interval length suitable for analysis on a time axis relating to an event is provided, and the corresponding definition value is set as a time interval length for an event to be analyzed. Analysis processing device.

Images