JP2012043113A - Program and medical examination support device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to provide a user (an analyst) with a group of medical examination items adapted to the medical examination history of an actual analysis object as a candidate for the group of medical examination items treated as analytical set.SOLUTION: An extraction node candidate generation part 104 analyzes a medical examination history record group in a medical examination history DB 100 and extracts a combination of medical examination items, the combination in which the number of applied patients is equal to or more than a threshold value, from among the combinations of medical examination items included in the records. Then a combination in which medical examination items are not the same classification is excluded from among the extracted combinations by referring to a classification DB 102. An extraction node selection reception part 108 presents remaining combinations as extraction node candidates to an analyst and receives selection of actually used abstract nodes among them. An abstraction processing part 112 converts the medical examination items of each medical examination history record into corresponding abstract nodes. A pattern mining part 116 performs pattern mining by using the record after the conversion.

Description

  The present invention relates to a program and a medical assistance device.

  A technique for generating a standard medical treatment process (clinical path) related to a medical condition from a history of medical care activities (for example, examination, medication, injection, etc.) performed on a patient with a medical condition in the past is known. . The clinical path is obtained, for example, by extracting medical items applied to many patients under the same condition (for example, the same disease) in the past medical history.

  In the system described in Patent Document 1, medical practice data (which may have a hierarchical structure) is stored in advance in medical data storage means. This system outputs information obtained by organizing a plurality of medical practice data into a graph format in a predetermined format. In this system, medical data is stored in a pre-designed framework, and the data is aggregated or graphed according to a specific format to support the planning of a medical plan.

JP 09-185651 A

  Since there are many medical care items (for example, specific drugs to be administered), there are cases where it is desired to collect a plurality of medical care items into one group and analyze the medical care process in units of groups. For example, there may be multiple medical items that can be used for the same purpose and can be substituted for each other. In such a case, such medical items of the same type rather than the individual medical items applied themselves. The application of may be important for analysis.

  In a system in which a history of medical care items is recorded for the purpose of analyzing a medical care process, information on a framework indicating which group each medical care item belongs to may be defined. For this reason, it is possible to analyze the accumulated medical history information according to the prepared framework and to extract the clinical path according to the analysis.

  By the way, when analyzing medical history information accumulated without assuming a medical process analysis such as medical history accumulated for medical fee billing, There are cases where analysis is performed along the framework. In such cases, the analyst needs to define the grouping framework because there is no information about the grouping framework or because the existing framework is not used. Here, since the number of medical care items is large, defining a combination of medical care items from the viewpoint of analysis from among them may be a heavy burden for an analyst (for example, a doctor). Although it is conceivable to analyze the raw medical items without grouping them, it is not possible to obtain an analysis result that can be understood for the first time when they are grouped together. For example, when an analysis is performed in units of individual medical treatment items, there are cases where a pattern that is cut off and not extracted at the time of analysis occurs because the number of corresponding medical history is small.

  It is an object of the present invention to provide a mechanism for providing a user (analyzer) with a medical history group to be treated as a group for analysis as a candidate for a group of medical treatment items in accordance with an actual analysis history group to be analyzed. And

  According to the first aspect of the present invention, there is provided a computer for determining the number of applied patients from a plurality of medical history records including at least patient identification information for identifying a patient and information indicating medical items applied to the patient. Extraction means for extracting a combination of medical treatment items whose total is equal to or greater than a predetermined threshold, and presenting the combination of medical treatment items extracted by the extraction means as extraction item candidates, and abstracting among the abstract item candidates from the user Selection acceptance means for accepting selection of items to be adopted as items, replacement means for replacing each medical item in each medical history record with an abstract item corresponding to the medical item, and medical history record in which the medical item is replaced with a corresponding abstract item Program to function as a pattern mining means for pattern mining It is.

  The invention according to claim 2 is the program according to claim 1, wherein the extraction means uses a minimum support value used by the pattern mining means in the pattern mining as the threshold.

  The invention according to claim 3 is such that the extraction means is a combination of medical items in which the total number of applied patients is equal to or greater than a predetermined threshold, and the medical items included in the combination are not of the same classification The program according to claim 1, wherein: is not extracted.

  According to a fourth aspect of the present invention, each of the medical history records includes information on a medical treatment category, and the extraction unit is configured to determine the number of applied patients among the medical history record groups corresponding to the same medical treatment category. The program according to any one of claims 1 to 3, wherein a combination of medical treatment items whose total is equal to or greater than a predetermined threshold is extracted.

  In the invention according to claim 5, the total number of applied patients is previously calculated from a plurality of medical history records including at least patient identification information for identifying a patient and information indicating medical items applied to the patient. Extraction means for extracting a combination of medical items that are equal to or greater than a predetermined threshold, and combinations of medical care items extracted by the extraction means are presented as extraction item candidates, respectively, and from the abstract item candidates as abstract items Selection accepting means for accepting selection of what to adopt, replacement means for replacing each medical item in each medical history record with an abstract item corresponding to the medical item, and a medical history record in which the medical item is replaced with a corresponding abstract item And a pattern mining means for performing pattern mining on the subject.

  According to the invention according to claim 1 or 5, an abstract item candidate that is a combination of medical items treated as a unit for analysis is selected as an abstract item candidate according to a medical history group to be actually analyzed, to a user (analyzer). Can be provided.

  According to the second aspect of the present invention, combinations that are not extracted in pattern mining can be prevented from being presented to the user as extraction item candidates.

  According to the invention of claim 3, extraction item candidates to be presented to the user can be narrowed down according to the classification.

  According to the invention which concerns on Claim 4, the extraction item candidate shown to a user can be narrowed down according to a medical treatment division.

It is a block diagram which shows the example of the outline of a structure of a medical treatment assistance apparatus. It is a figure which shows the example of the data content of medical treatment history DB. It is a figure which shows the example of the data content of classification DB. It is a flowchart which shows the example of the procedure of the process which a medical treatment assistance apparatus performs. It is a figure which shows the example of bit pattern data. It is a figure for demonstrating the number of patients applicable to one specific medical treatment item in bit pattern data. It is a figure for demonstrating the number of patients applicable to either of two specific medical treatment items in bit pattern data. It is a figure which shows the example of the obtained extraction node candidate group. It is a figure which shows the example of the definition information of the selected extraction node. It is a figure which shows the example of the tree structure which represents a medical treatment log | history typically. It is a figure which shows the example of the result of having substituted the concrete medical treatment item of the medical treatment history illustrated in FIG. 10 by the abstract node. It is a figure which shows the example of the common pattern calculated | required from the medical treatment history group after the abstract node replacement illustrated in FIG. It is a figure which shows the example of the data content of the medical treatment history DB in a modification. It is a flowchart which shows the example of the procedure of the process which the medical assistance device in a modification performs.

  A medical assistance device according to an example of an embodiment of the present invention is an apparatus that supports analysis of a history of medical treatment processes actually performed on a patient. The medical process means a set of medical activities performed in the treatment of a patient. Hereinafter, individual medical treatments constituting the medical care process from the viewpoint of data processing are referred to as medical care items. The clinical items are conceptually based on specific levels such as the administration of specific drugs identified by product names and generic names, specific injections, injections of specific liquids, and gallbladder ultrasonography. Can be captured at various stages up to various levels (classification levels such as examination, medication, and injection). However, here, it is assumed that medical history data in which individual medical items are recorded at a specific level is given. The medical assistance device according to the present embodiment supports the work of organizing and analyzing these specific medical treatment items from the viewpoint of analysis desired by an analyst (for example, a doctor). For example, individual specific medicines “A” and “B” are handled in one group from the commonality of efficacy, and for example, a medical history using the group (any medicine belonging to) is found out. Supports analysis such as extracting frequent patterns of medical history (pattern mining) using groups instead of specific medical items as a unit.

  In order to handle a plurality of medical treatment items as a group from the viewpoint of analysis, for example, it is considered to create a rule for grouping in advance and automatically group medical treatment items according to the rule. It is done. This method is effective when the viewpoint of classification is fixed, but cannot be applied when the viewpoint of analysis changes according to the purpose of the analyst (such as a doctor). For example, when the analyst considers the drugs “A” and “B” as one group with respect to the same medical history data, and analyzes the drugs “A”, “B”, and “C” as one group In some cases, it may be desirable to perform both analyzes at the same time or step by step. It is also conceivable that various collective rules are prepared in advance in anticipation of various needs, and the analyst selects a rule that matches his / her purpose. However, for example, even in one disease of “stomach cancer”, there are more than 80 types of medical treatment items applied to the treatment. For example, even if the 80 sets of two or three sets are grouped, the combination is about 100,000. A huge number. It is not realistic for an analyst to select one that meets his needs from such a huge amount.

  Therefore, the medical assistance device of this embodiment analyzes the medical history data that is the target of the actual pattern analysis, finds a combination of meaningful medical items that are actually grouped in the medical history data, Take an approach that lets analysts select what they want to use for analysis. A meaningful combination (group) collectively in a group is a combination (group) that appears sufficiently in the analysis history data to be analyzed (that is, appears in a number on a predetermined threshold). In particular, in this embodiment, by using the minimum support value specified in pattern mining as a threshold value, combinations that cannot be extracted by pattern mining are omitted from the options in advance.

  Details of this embodiment will be described below. Hereinafter, a group in which one or more medical items are collected is referred to as an “abstract node”.

  FIG. 1 shows an example of a schematic internal configuration of a medical assistance device. The medical assistance device 10 includes a medical history DB (database) 100, a classification DB 102, an abstract node candidate generation unit 104, an extraction condition storage unit 106, an abstract node selection reception unit 108, an abstract node definition storage unit 110, an abstraction processing unit 112, An abstraction history DB 114, a pattern mining unit 116, an output processing unit 118, an input unit 120, and a display unit 122 are provided.

  The input unit 120 receives an input such as an instruction by a user (analyzer) to the medical assistance device 10. The input unit 120 may be realized by an input device such as a keyboard, a mouse, or a touch panel. The input unit 120, for example, a narrowing-down condition for narrowing down the medical history to be analyzed from data in the medical history DB 100 (hereinafter referred to as “population condition” in the sense of a condition for specifying the analysis population), pattern mining An input such as a minimum support value used in the unit 116 or the like is received. The input minimum support value is stored in the extraction condition storage unit 106.

  In pattern mining, items whose appearance frequency is less than or equal to the minimum support value do not affect the mining result (that is, they do not appear in the mining result). That is, the minimum support value is a threshold value that serves as a basis for cut-off in mining. The minimum support value is specified, for example, as a percentage of the total number of data (patients) in the population or the absolute number of data (patients). For example, in a case where a patient group consisting of 46 patients whose main disease name is “stomach cancer” is a population, if the minimum support value is set to “90%”, the pattern corresponding to 41 or more patients is included in the mining result. Only appears. The same applies when the minimum support value is set to “41 people”. The minimum support value may be determined according to the conditions under which the analyst wants to extract and view the pattern. To obtain an outline of a certain disease (for example, a typical medical process), a designation such as “90%” can be considered. Or if you want to see a relatively exhaustive pattern without limiting it to a typical one, but want to discard a pattern that can only be seen with too few people (for example, less than 10 people), Such a specification can be considered. Thus, the analyst inputs the minimum support value from the input unit 120 according to the purpose of his / her analysis.

  The display unit 122 displays the result of processing in the medical assistance device 10. Display according to the display control information from the output processing unit 118 is performed. The display unit 122 is realized by a display device such as a liquid crystal display, for example.

  The medical history DB 100 is a storage unit that stores information on the history of medical processes performed on patients. In this example, the medical history DB 100 stores a medical practice performed from one hospitalization to discharge of one patient and the time of implementation as a single medical history. The medical history DB 100 is configured such that, for example, a management device (not shown) receives input of information related to a medical process by doctors, nurses, laboratory technicians, and other staff working at medical institutions, and registers them in the medical history DB 100. Built.

  FIG. 2 shows an example of the data contents of the medical history DB 100. In the example of FIG. 2, information on the main disease name, medical treatment item, and implementation time is included in association with a patient ID (identifier) that is identification information uniquely assigned to each patient. The “main disease name” represents the name of the main disease to be treated by the patient in the hospitalization, and is recorded, for example, for calculating a medical fee. The “medical item” is information representing a specific medical practice applied to the patient. For example, when a medical practice of administering “L” (a drug product name) is performed, the name “L” of the drug is recorded in the column of the medical treatment item. The “implementation time” represents the time when the medical treatment indicated by the “medical item” is performed. In FIG. 2, values based on the date of surgery, such as “the day before surgery”, “the day of surgery”, “the day after surgery”, “the second day after surgery”, and the like are used as the values for the implementation time. In the table of FIG. 2, one line including the patient ID, the main disease name, the medical treatment item, and the implementation time is hereinafter referred to as a medical history record. The medical record record may include information items other than the exemplified patient ID, main disease name, medical items, and implementation time. The history of a medical process in one hospitalization of one patient is expressed as a set of one or more medical history records.

  The classification DB 102 is a database that stores medical item classification information. As for the medical treatment item corresponding to the medication, for example, the classification DB 102 that stores classification information as shown in FIG. 3 may be used. In the classification information of FIG. 3, for each specific drug name (such as a product name or a general name), information on the major classification and middle classification to which the medicine belongs is registered. From the example of FIG. 3, it can be seen that, for example, the drug “A” belongs to the middle classification “preparation for acidosis” in the large classification “infusion / nutritional preparation”. Instead of constructing and using the classification DB 102, for example, an existing drug DB that is publicly available (however, having information on drug classification) may be used.

  The abstract node candidate generation unit 104 is a functional module that analyzes the medical history information to be analyzed in the medical history DB 100 and generates abstract node (medical item group) candidates. The abstract node candidate generation unit 104 generates an abstract node candidate with reference to the minimum support value and the population condition input from the input unit 120 (stored in the extraction condition storage unit 106).

  The abstract node selection accepting unit 108 displays the information of each abstract node candidate generated by the abstract node candidate generating unit 104 on the display unit 122, for example, and is actually used from among these candidates from the analyst via the input unit 120. It accepts selection of abstract nodes. Information on each selected abstract node, for example, information on a combination of medical items constituting each abstract node, is stored in the abstract node definition storage unit 110.

  The abstraction processing unit 112 performs a process of replacing each medical item included in each medical history record belonging to the population with the selected abstract node. The medical history record in which the medical items are replaced with the corresponding abstract nodes in this way is hereinafter referred to as an abstract history record. The abstraction history record generated by the abstraction processing unit 112 is accumulated in the abstraction history DB 114.

  An example of a processing procedure performed by the abstract node candidate generation unit 104, the abstract node selection reception unit 108, and the abstraction processing unit 112 will be described with reference to FIG.

  In this procedure, first, the abstract node candidate generation unit 104 acquires a population condition and an extraction condition (minimum support value) (S10). These are preliminarily input from the user (analyzer) as conditions for pattern mining.

  The population condition is a condition for narrowing down the medical history to be analyzed, and is specified as a search condition for each data item included in the medical history. For example, if it is desired to use the treatment history of patients whose main disease is “stomach cancer” as the analysis population, the population condition “main disease name =“ stomach cancer ”” may be used. In addition, when it is desired to use the medical history of patients to whom the drug “A” is administered as a population, the population condition of “medical treatment item =“ A ”” may be used. Of course, a logical expression of conditions for a plurality of data items may be used as the population condition.

  The abstract node candidate generation unit 104 extracts a medical history record that satisfies the population condition from the medical history DB 100 (S12). In this extraction process, a medical history record that satisfies the population condition may be simply extracted, or a medical history record that corresponds to the same hospitalization of the same patient as the medical history record that satisfies the population condition may be extracted. The extracted record group becomes the analysis population.

  Next, the abstract node candidate generation unit 104 analyzes the extracted populations and obtains medical items whose number of corresponding patients is equal to or greater than the minimum support value by combining them. For this processing, in this embodiment, bit pattern data is generated from the population (S14).

  FIG. 5 shows an example of bit pattern data. In this example, each row represents an individual patient and each column represents an individual medical item. A cell specified by a combination of a row and a column is a binary bit value indicating whether or not a medical treatment item corresponding to the column is applied to a patient corresponding to the row. That is, the value of the bit is set to “1” if it has been applied, and is set to the value “0” if it has not been applied.

  In S14, the abstract node candidate generation unit 104 first generates bit pattern data in which all bits (cells) are set to “0”. Then, the patient ID and the value of the medical item in each medical history record in the population are examined, and when the record indicating that a certain medical item is applied to a certain patient is found, the patient and the medical treatment in the bit pattern data are detected. The bit corresponding to the combination with the item is set to “1”. If the bit corresponding to the combination of the patient and the medical item included in the medical history record is already set to “1”, the bit remains “1”. The bit pattern data is completed by performing this process for all the medical history records in the population.

  Next, the abstract node candidate generation unit 104 extracts a combination of medical items satisfying the extraction condition from the completed bit pattern data (S16). In this example, the extraction condition is the minimum support value. That is, for each combination of one or more medical items, the number of patients whose bit corresponding to one or more medical items belonging to the combination is “1” is totaled, and the number is equal to or greater than the minimum support value. Extract combinations. In this extraction process, an upper limit is set for the number of medical treatment items to be combined, and only combinations including medical treatment items equal to or less than the upper limit number are determined. For example, when a combination of up to three medical items is targeted, a combination consisting of one medical item (medical item itself), a combination consisting of two, and a combination consisting of three are targeted.

  The process of S16 will be described using a specific example. Consider an example in which the minimum support value is set to 5 for the bit pattern data whose population is the medical history of 10 patients illustrated in FIG. In this case, considering only a single medical item “A” as illustrated in FIG. 6, the total number of patients corresponding to the medical item (that is, the bit value is “1”) is three. Yes, less than the minimum supported value. Therefore, a combination including one medical treatment item “A” is not extracted in S16. On the other hand, in the combination of two medical treatment items “A” and “B” as illustrated in FIG. 7, the total number of patients corresponding to at least one of the two is five, which is equal to or greater than the minimum support value. Therefore, in this example, the combination of the medical treatment items “A” and “B” is extracted in S16.

  Thus, in S16, for each combination consisting of the number of medical items equal to or less than the upper limit number, the total number of patients corresponding to at least one medical item in the combination is obtained, and if the total number satisfies the extraction condition, That is, if the minimum support value is exceeded, the combination is extracted.

  Here, in the present embodiment, the abstract node candidate generation unit 104 does not use the extracted combination of medical items as such as a candidate for an abstract node as it is, but refers to the classification DB 102 so that the medical items that are not the same item are referred to. A combination of items is excluded (S18). For example, a combination consisting of the three treatment items “K” and “L” (see FIG. 3) belonging to “vitamin preparation” and the treatment item “Q” belonging to “peptic ulcer drug” was extracted in S16. However, since the classification of these three items is not the same, this combination is excluded in S18.

  As shown in FIG. 3, since there are a plurality of levels for classification, the analyzer may be able to set for the medical assistance apparatus 10 at which level the identity of the classification is determined in S18. . For example, in the example of FIG. 3, the medical treatment items “K” and “M” belong to the same “vitamin preparation” in the large classification, but differ in the middle classification as “mixed vitamin B group” and “pantothenic acid”. Become a classification. Therefore, the classification level for determining the identity may be specified according to the analysis intention of the analyst.

  Instead of this, the abstract node candidate generation unit 104 stores the classification levels determined as the same classification, and when the abstract node candidates are presented later, at which classification level the medical treatment items constituting the candidates are displayed. You may display whether it was determined with the same classification.

  In S18, the combination of the medical items that do not belong to the same classification is excluded by sifting the combination of the medical items in accordance with the classification system registered in the classification DB 102. The medical items belonging to the same classification in the classification system of the classification DB 102 are highly likely to be meaningful in a group (abstract node), and conversely, items that do not belong to the same classification are meaningless. Since it is likely, such a sieve reduces the possibility of presenting a meaningless option to the analyst.

  The abstract node candidate generation unit 104 passes the combinations that remain without being excluded in the sieving process based on the classification DB 102 to the abstract node selection reception unit 108 as abstract node candidates (S20). At this time, each abstract node candidate may be passed in association with information on the classification to which the medical care item belonging to the candidate belongs. For example, when the two-level classification system of major classification and middle classification as shown in FIG. 3 is adopted, abstract node selection is performed for each corresponding classification name of major classification and middle classification for each medical item belonging to the candidate. For example, it is passed to the reception unit 108. Of course, this is only an example. For example, medical items with the same middle classification also have the same major classification. In such a case, only the classification name of the middle classification may be associated with the candidate. In addition, for candidates consisting of medical items that do not match the middle classification but match the major classification, only the classification name of the major classification may be associated with the candidate.

  The abstract node selection reception unit 108 displays each abstract node candidate received from the abstract node candidate generation unit 104 as an option on the display unit 122 via the output processing unit 118 in the form of, for example, a list display (of course, the display form is Not limited to this). Then, from among the abstract node candidates, selection of what the analyst uses in pattern mining is accepted (S22).

  FIG. 8 shows an example in which eight abstract node candidates (a) to (h) are obtained as options from a certain population. In this example, each candidate (each row in the table of FIG. 8) includes information on the medical item belonging to the candidate and the classification name to which the medical item belongs. In this example, the identity of the classification is determined based only on the large classification in the classification system of FIG. Such a list is displayed on the display unit 122 as an analyst's choice. In addition, when the identity determination is performed in two stages of the large classification and the medium classification, for example, in the display of the classification name of each medical treatment item of the abstract node candidate, the classification level determined to be the same classification can be identified. Such display may be performed. For example, the background or character display color in the category name column is changed depending on whether the major classification is the same classification or the middle classification. In addition, even when it is determined that the major classification is the same, not only the major classification but also the middle classification may be displayed. The display of the corresponding classification for each candidate or the individual medical items belonging to it becomes a judgment material when the analyst makes a selection with respect to the abstract node candidate. In the example of FIG. 8, the candidates (a) to (c) are composed of a single medical item, but (d) to (h) are combinations of a plurality of medical items.

  Note that the classification of the second medical treatment item “M” of the candidate (d) illustrated in FIG. 8 is “no corresponding classification”. This means that the classification information of the medical treatment item “M” is not registered in the classification DB 102. That is, in the example of FIG. 8, a combination of the medical item “A” corresponding to the classification “vitamin preparation (injection)” and the medical item “M” “not applicable” is extracted as an abstract node candidate (d). Has been. As described above, in the present embodiment, the medical item “No applicable classification” does not affect the determination of whether the medical items constituting the combination are of the same classification. That is, if the medical items other than the medical item “No applicable classification” included in the combination belong to the same class, the combination is determined to be the same class.

  The abstract node selection receiving unit 108 displays abstract node candidate information as illustrated in FIG. 8 on the display unit 122 and receives a selection from the analyst via the input unit 120. The analyst refers to the combination of medical items belonging to each abstract node candidate, and if necessary, refers to the information such as the classification corresponding to the combination or the classification to which each medical item belongs, An abstract node candidate that matches the purpose is selected as a formal abstract node by an operation of the input unit 120 (for example, an operation of selecting a desired candidate using a pointing device). The abstract node selection receiving unit 108 receives the selection result (S22). The abstract node selection receiving unit 108 receives an input of a label to be given to each selected abstract node from the analyst via the input unit 120. This label is an identifier given by the analyst to identify individual abstract nodes. When the analyst does not input a label to be assigned to the abstract node, a classification name that matches between the medical items belonging to the abstract node may be adopted as a label of the abstract node as a default value.

  The abstract node selection receiving unit 108 stores the definition information of each abstract node selected and labeled in this way in the abstract node definition storage unit 110. For example, as shown in FIG. 9, the definition information of each abstract node includes information about a label assigned to the abstract node and a list of specific medical care items belonging to the abstract node. In addition, information such as a classification corresponding to the abstract node, a classification to which each medical item belonging to the abstract node belongs, or a level of the classification may be included.

  In this way, when an abstract node to be used for analysis is selected by the analyst, the abstraction processing unit 112 refers to the definition information of these abstract nodes stored in the abstract node definition storage unit 110 and refers to the medical history DB 100. Each medical history record of the population is abstracted (S24). In this abstraction, for example, for each medical history record, an abstract node that includes the medical item in the record in the medical item list is searched from the abstract node definition storage unit 110. If such an abstract node is found, Replace the value of the medical item with the label of the abstract node. When the abstract node corresponding to the medical item in the medical history record is not found, such replacement may not be performed.

  For example, it is assumed that medical process histories 200 and 202 for patients “001” and “005” illustrated in FIG. 10 are included in the population. The illustrated medical process history schematically shows a set of a plurality of medical history records for the same patient (a medical history record is a record for one medical item) in a tree display. When abstraction processing to the abstraction node in FIG. 9 is performed on these medical process histories, abstraction process histories 200A and 202A as illustrated in FIG. 11 are obtained. The abstract process history is a collection of a plurality of abstract history records for the same patient.

  The medical history record after this replacement (referred to as an abstract history record) is stored in the abstract history DB 114.

  When all the medical history records in the population are abstracted, the pattern mining unit 116 performs pattern mining processing on the abstract history record groups stored in the abstract history DB 114. For example, the pattern mining unit 116 is a tree mining technique (a reference document is, for example, M. Zaki. Efficiently mining frequent trees in a forest. 8th ACM SIGKDD Int'l Conf Knowledge Discovery and Data Mining, July , 2002.) or the like is used to extract a pattern including a common abstract node or a common medical care item. For example, from the abstraction process histories 200A and 202A of FIG. 11, as illustrated in FIG. 12, a common pattern 300 indicating that “antibiotics and vitamin preparations are administered on the day of surgery” (“clinical history pattern # 1”) Is extracted.

  In this pattern mining process, the minimum support value stored in the extraction condition storage unit 106 is used as the minimum support value. That is, the pattern mining unit 116 extracts only patterns having a support value equal to or greater than the minimum support value. The minimum support value used here is the same as the extraction condition used by the abstract node candidate generation unit 104 to cut off the abstract node candidate.

  Further, the pattern mining unit 116 obtains the number of abstraction histories having a pattern that matches the common pattern for each extracted common pattern with reference to the abstraction history DB 114, and obtains a score corresponding to the number. Also good. Then, the extracted common patterns may be sorted and displayed in the order of scores, or the scores may be displayed in association with each common pattern. Note that the method for obtaining the score of the common pattern is not limited to the above. Instead, for example, the method disclosed in Japanese Patent Application No. 2010-149691 (filed on June 30, 2010) by the applicant, that is, a clinical path for each common pattern (standard practice for the treatment of individual diseases) The degree of coincidence with the process) may be obtained as a score.

  The output processing unit 118 generates a display image representing the result of pattern mining and displays it on the display unit 122. The display form may be a tree shape as illustrated in FIG. 12, or may be another form.

  As described above, in this embodiment, a rule for abstraction (that is, abstract node definition) when a plurality of medical treatment items are abstracted (grouped) in pattern mining is an object of pattern mining. Obtained based on the population of actual medical history records. In the present embodiment, only combinations of medical items that are abstracted and meaningful (that is, appear as many times as they are extracted without being cut off in pattern mining) are extracted as candidates, and an actual analysis is performed from these candidates to the analyst. Let them choose what to use. Conversely, it is known that a combination of medical items that appears so few that they cannot be extracted at the time of pattern mining is not extracted as a pattern even if it is an abstract node. The smaller the number of candidates, the less the burden on the analyst's selection work.

  In the embodiment described above, the combinations of the medical items extracted in S16 are narrowed down with reference to the classification DB 102 in S18, but such narrowing is not necessarily essential. In the present embodiment, combinations whose support values are equal to or greater than the minimum support value are extracted in S16, and the number of combinations extracted in this extraction process is sufficiently small in view of the burden on the analyst's selection work (for example, There are cases where the number is equal to or less than a certain threshold value.

  Next, a modification will be described with reference to FIGS. 13 and 14. In this modified example, as shown in FIG. 13, information of “medical treatment classification” included in the medical history record is used. “Medical classification” is information indicating a classification of a situation or the like to which a medical treatment item in the medical history record is applied. For example, the “2010 DPC Introduction” by the Ministry of Health, Labor and Welfare's Insurance Bureau's Medical Division published at the URL “http://www.mhlw.go.jp/topics/2010/06/dl/tp0610-1a.pdf” P120-132 “Reception data collection by the receipt data download method related to“ Evaluation of DPC impact ”” in p. E-5, F-5, EF-5) ”may be used as the“ medical division ”in this modification. This data classification includes, for example, types of medication such as “internal use”, “free use”, “external use”, and types of injection such as “subcutaneous intramuscular” and “intravenous”. The classification information illustrated in FIG. 3 is a classification based on a drug type (drug effect), but this “medical treatment classification” may be regarded as a classification code from a different viewpoint.

  In this modification, the extraction node candidate generation unit 104 does not select combinations of medical care items with different medical care classifications as abstract node candidates. That is, in this modified example, if the medical treatment classification is different, the situation where the medical treatment item is applied is different, and thus the medical treatment items having different application situations are not abstracted.

  An example of the processing procedure of this modification is shown in FIG. 14, steps corresponding to the same steps as the steps in the procedure shown in FIG. 4 are denoted by the same reference numerals and description thereof is omitted.

  In the procedure of FIG. 14, after abstracting the population (S12), the abstract node candidate generation unit 104 extracts the medical history record corresponding to the medical treatment category from the population (S13) and extracts the medical treatment records. Abstract node candidates are extracted by performing the processing of S14 to S20 for the record group. The processes in S13 to S20 are repeated for each medical treatment category. In this way, by extracting abstract node candidates from only the records corresponding to the medical care category for each medical care category, combinations of medical care items across the medical care categories are prevented from becoming abstract node candidates. Processing S22 and S24 after the abstract node candidate is obtained in this way may be the same as the example of FIG.

  In this modified example, combinations of medical care items that are abstract node candidates are screened by medical care classification. In the procedure of FIG. 14, the abstract node candidates are narrowed down using both the screening based on the medical classification and the screening based on the medicinal effect classification in the classification DB 102 (S 18), but a method using only the medical classification is also conceivable.

  In the above embodiment and modification, an abstract node definition (see FIG. 9) is created and stored in the abstract node definition storage unit 110. This abstract node definition is saved so that it can be reused in later analysis. May be. For example, once created abstract node definitions may be stored as dictionary data and used to narrow down combinations of medical items, as with the classification information in the classification DB 102. That is, in this case, even if the combination of the medical items extracted in S16 is not the same classification in the classification system of the classification DB 102, if it is included in the combination of the medical items in any abstract node in the dictionary data, Leave the combination without excluding it. That is, the abstract node definition dictionary data is used as classification information unique to the analyst. When a plurality of classification systems are registered in the classification DB 102 and the analyst selects and uses the plurality of classification systems in accordance with the analysis intention at the time of individual analysis, the classification that is leaked from the selected existing classification system is selected. It can be used such as covering with a dictionary of past abstract node definitions. Further, if personal dictionary data is enriched, S18 may be narrowed down using only the dictionary data without using the classification DB.

  The medical assistance device 10 exemplified above is realized, for example, by causing a general-purpose computer to execute a program representing the processing of each functional module described above. Here, the computer includes, as hardware, a microprocessor such as a CPU, a memory (primary storage) such as a random access memory (RAM) and a read only memory (ROM), an HDD controller that controls an HDD (hard disk drive), Various I / O (input / output) interfaces, network interfaces that perform control for connection to a network such as a local area network, and the like have a circuit configuration connected via a bus, for example. Also, portable non-volatile recording of various standards such as a disk drive and a flash memory for reading and / or writing to a portable disk recording medium such as a CD or a DVD via the I / O interface, for example. A memory reader / writer for reading from and / or writing to a medium may be connected. A program in which the processing contents of each functional module exemplified above are described is stored in a fixed storage device such as a hard disk drive via a recording medium such as a CD or DVD, or via a communication means such as a network, and stored in a computer. Installed. The program stored in the fixed storage device is read into the RAM and executed by a microprocessor such as a CPU, thereby realizing the functional module group exemplified above.

  DESCRIPTION OF SYMBOLS 10 Medical treatment support apparatus, 100 Medical treatment history DB, 102 Classification DB, 104 Abstract node candidate production | generation part, 106 Extraction condition memory | storage part, 108 Abstract node selection reception part, 110 Abstract node definition memory | storage part, 112 Abstraction processing part, 114 Abstraction History DB, 116 pattern mining unit, 118 output processing unit, 120 input unit, 122 display unit.

Claims (5)

Computer
Medical items whose total number of applied patients is equal to or greater than a predetermined threshold from a plurality of medical history records including at least patient identification information for identifying patients and information indicating medical items applied to the patients Extraction means for extracting a combination of
A selection receiving unit that presents each combination of medical items extracted by the extraction unit as an extraction item candidate, and receives a selection of a candidate to be adopted as an abstract item from the abstract item candidates;
Replacement means for replacing each medical item in each medical history record with an abstract item corresponding to the medical item,
Pattern mining means to perform pattern mining for medical record records that replace the medical items with corresponding abstract items,
Program to function as.   The program according to claim 1, wherein the extraction unit uses a minimum support value used by the pattern mining unit in the pattern mining as the threshold. Even if the extraction means is a combination of medical items in which the total number of applied patients is equal to or greater than a predetermined threshold, the medical items included in the combination are not extracted if they are not in the same classification.
The program according to claim 1 or 2, characterized by the above-mentioned. Each of the medical history records includes information on the medical category,
The extraction means extracts a combination of medical items in which the total number of applied patients is equal to or greater than a predetermined threshold in a medical history record group corresponding to the same medical classification,
The program according to any one of claims 1 to 3, wherein: Medical items whose total number of applied patients is equal to or greater than a predetermined threshold from a plurality of medical history records including at least patient identification information for identifying patients and information indicating medical items applied to the patients Extraction means for extracting a combination of
A selection receiving unit that presents each combination of medical items extracted by the extraction unit as an extraction item candidate, and receives a selection of a candidate to be adopted as an abstract item from the abstract item candidates;
A replacement means for replacing each medical item in each medical history record with an abstract item corresponding to the medical item;
Pattern mining means for pattern mining for medical history records in which medical items are replaced with corresponding abstract items;
A medical assistance device comprising:

Images