JP4032775B2 - Encoding system and program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a coding processing technology that facilitates coding of words with fluctuations in notation, in particular, an electronic medical record system for managing the names of wounds and diseases for medical institutions to analyze disease trends, etc., or insurers and examinations The present invention relates to a receipt processing system in which a payment institution encodes an injury / illness name described in a medical fee description and performs statistical processing.
[0002]
[Prior art]
When a doctor examines a patient, he or she records the name of the patient's injury based on the medical condition and condition. The notation method of injury and illness names is left to the discretion of individual doctors, and in addition to the standard injury and illness names, there are various injury and illness names to express the same medical condition and pathology, such as different idiomatic injury and illness names and kana names. appear. In addition, various modifiers that more accurately describe the patient's condition and medical condition are added, so the notation of wound names is more diverse, and these notation fluctuations are a major obstacle from the viewpoint of code management of wound names. ing.
For example, an insurer such as a health insurance association aggregates the names of injuries and sicknesses listed in the injuries and sickness names column of the medical treatment remuneration statement (hereinafter referred to as “receipt”) issued by each medical institution to effectively develop health services. Statistical processing of disease trends is performed for the purpose.
In order to analyze the disease trend, it is necessary to encode and summarize the names of wounds, but conventionally this work is done by manually converting the names of wounds and diseases described in the receipt into codes corresponding to experienced workers, I was counting. However, some insurers receive hundreds of thousands to millions of receipts per month, and it is desired that the monthly summation is heavy and burdensome in terms of time and cost.
In order to realize the mechanization of reading the name of sickness by the conventional technique, for example, the character string of the name of the sickness described in the receipt is input by an OCR device (character reading device) or punch input, and the recording medium storing the name of the sickness and the disease is stored. There is a method of coding a corresponding disease name by collating with a certain disease name master. However, since the injury / illness name character string described in the receipt has the above-described fluctuation, it is not always easy to reliably obtain a code even if it is checked against the injury / illness name master. For this reason, it is necessary to provide a user dictionary separately from the disease name master, and to register the disease name character strings described in the receipt that have not been detected together with the corresponding codes.
[0003]
[Problems to be solved by the invention]
In order to automatically code many names of injuries and diseases according to the conventional technique, it is necessary to register more notation fluctuation names in the user dictionary.
For this reason, it has been necessary to visually check the names of wounds and diseases that could not be coded at the time of collation of the names of the disease and disease by the conventional technique, and add them to the user dictionary as necessary.
In the registration work in the user dictionary, the wound name on the wound name master that seems to be the closest to the wound name that could not be coded is searched for and associated with the code. This operation is time-consuming and may cause variations and registration errors depending on the registered user.
In addition, when changing the code system used for code management, it is necessary to re-associate the codes with all the names of wounds in the user dictionary by using a new code system.
In this way, coding techniques using user dictionaries impose a great deal of effort on the part of the user, so code management using only the standard injury name master as much as possible improves the work efficiency as well as variations and registration errors. It can be said that it is safe from the viewpoint of prevention.
Accordingly, an object of the present invention is to provide a coding method that can easily collate more notational fluctuations with a disease name master with less burden.
Another object of the present invention is to provide an environment for evaluating the wound name and the degree of matching that are considered appropriate and presenting them to the user when there is no perfectly matched wound name as a result of the wound name master comparison. It is to provide.
Another object of the present invention is to present a plurality of wound names on the wound name master that is considered to be approximate when the user associates with a code of another wound name as a result of the evaluation, so that the user can obtain the optimal wound name. It is to provide a means for selecting.
[0004]
[Means for Solving the Problems]
In order to achieve the above object, the method of encoding a disease name according to the present invention provides:
Before collating the entered wound name with the wound name master, the notation fluctuation part on the wound name is converted into a standard notation so that the master wound name is easily detected.
In addition, in order to achieve the above object, the method of encoding a disease name according to the present invention,
When multiple wound names and modifiers are included in the input disease name string, the structure of them is analyzed and included in the wound name string based on the result. It is shown to the user whether or not encoding is possible using the name of the wound or illness.
In addition, in order to achieve the above object, the method of encoding a disease name according to the present invention,
When the user determines that another code should be associated with the input character string of the disease name, the user detects a plurality of approximate disease names of the input disease name and presents them to the user. Lets you select a code.
[0005]
DETAILED DESCRIPTION OF THE INVENTION
The encoding system which is the first embodiment of the encoding method of the present invention will be described below.
FIG. 1 is a basic configuration diagram of this embodiment. A computer apparatus 1 stores a program for executing the encoding method of the present invention on a recording medium. The disease name character string data 2 input through an OCR device, punch input, an external recording medium such as a flexible disk, a network connection with another computer device, etc., not shown in the figure, is encoded in the computer device 1 As a result, it is output as injury / illness name encoding result data 3. The output of the wound name encoding result data 3 may be recorded on an external recording medium such as a flexible disk, output to another computer device via a network connection, or printed on a form by a printer.
The computer apparatus 1 includes a notation fluctuation standardization means 4, wound name master matching means 6, character string structure analysis means 8, approximate wound name search means 10, matching result correction means 12, and control means 14 as processing programs. In addition, the recording area includes a notation fluctuation table 5, a disease name master table 7, a structural formula table 9, an approximate disease name table 11, an encoded result storage table 13, and a collation table 15, and the operator manages and corrects the processing results. As a peripheral device, a display unit 16 such as a display and an input unit 17 such as a keyboard and a mouse are provided.
The general operation of this embodiment will be described below with reference to FIG.
First, the input character string, which is the disease name character string data 2 acquired from the outside, is collated with the notation fluctuation table 5 by the notation fluctuation standardization means 4. The notation fluctuation table 5 stores in advance fluctuation phrases of idiomatic expressions and kana words and standard phrases that are standard notations. The fluctuation standardization means 4 stores fluctuation phrases in the input character string from the notation fluctuation table 5. Detect and convert to the corresponding standard phrase.
In the example of FIG. 2, the input character string “right forearm contact dermatitis” is input as the disease name character string data 2. The phrase “skin” is sometimes written in the notation “skinfu”, and “skinfu” is registered in advance in the notation fluctuation table 5 as “skinfu” and “skin” as its standard phrase. It is. Therefore, the input character string “right forearm contact dermatitis” is converted into a standardized character string “right forearm contact dermatitis”. By using the notation fluctuation standardization means 4 and the notation fluctuation table 5 in this way, it is possible to convert idiomatic expressions contained in the input character string and notation fluctuations due to variations in kana usage into standard notations. it can. Therefore, an input character string with fluctuation in notation can be encoded as a standard character string. When the same effect is realized without using the fluctuation standardization means 4 and the notation fluctuation table 5 as in the prior art, the user dictionary includes the phrase “skin” such as “dermatitis” and “skin laceration”. It is necessary to register the words “skin fever” and “skin fracture”, which are rewritten as “skin fu”, with the code. There are more than 400 standard names of injuries with “skin” alone, and the user dictionary will become huge if all the notation fluctuations are registered. For this reason, the burden of registration work in the user dictionary and the verification time are remarkably increased, and the risk of coding conversion due to an input error at the time of registration is also increased.
Therefore, as in the coding method according to the present embodiment, it is more efficient to register not only the entire wound name in the user dictionary, but also to check the standard wound name master after converting only the notation fluctuation part. The risk of mistakes is reduced.
Next, the generated standardized character string is collated with the wound name master collating means 6 by the wound name master collating means 6. In the wound name master table 7, standard wound names and modifiers added to the wound names are registered as master words, and codes and phrase attributes corresponding to the respective master words are stored. The wound name master collating means 6 detects all the master words / phrases included in the standardized character string, and extracts codes and word attributes of the words / phrases.
The phrase attribute here is a parameter indicating the type of master phrase, and is used to classify whether the name is a wound name or a modifier, or what type of modifier is a modifier. In the example of FIG. 2, as a result of collating the standardized character string “right forearm contact dermatitis” with the wound name master table 7, master words “right”, “forearm”, “contact dermatitis” are detected. The attributes are “position” which is a modifier indicating position, “part” which is a modifier indicating a part of the body, and “wound and disease” indicating the name of the wound.
From the result of the master collation, it is possible to determine that the input character string of injury / illness name is the injury / illness name “contact dermatitis” including a modifier for the position and part, and the code is “6929196”. However, in general, when an input character string includes a wound name and a modifier, it cannot be determined that the wound name is a wound name corresponding to the input character string. For example, the input character string may be described as “dermatitis (contact)”. In this case, as a result of collation with the wound name master, the wound name “dermatitis” and “pathogenesis” “contact” It is detected as a modifier that indicates that it is impossible to detect the name of a consecutive “contact dermatitis”. Therefore, the code is “6869043” which means “dermatitis”. As described above, even if the disease name included in the disease name character string is encoded as it is, an appropriate encoding result may not always be obtained.
Therefore, in this embodiment, whether or not the coding result is valid by using the concept of the structural formula indicating how the master words such as the names of sicknesses and modifiers included in the input character string are arranged. To evaluate.
The character string structure analyzing means 8 and the structural formula table 9 are used for the coding evaluation by the structural formula. The character string structure analyzing means 8 pays attention to the word attribute among the collation results generated by the disease name master collating means 6, arranges the word attributes corresponding to the order of the master words detected in the disease name character string, Treat as. The character string structure analyzing means 8 collates this structural formula against the structural formula table 9, and when the same structural formula is detected, the code of the wound name included in the input wound name character string from the corresponding structural evaluation. Judge whether or not can be adopted. If the same structural formula is not found, it is determined that it cannot be adopted.
In the example of FIG. 2, in the case of “right forearm contact dermatitis”, the structural formula is “positional site injury”. This structural formula is registered as “good” in the structural formula table 9 as the structural evaluation, and therefore, it is determined that the code of the wound name “contact dermatitis” included in the wound name string can be adopted. To do.
Since there are an infinite number of combinations of injury and disease name strings composed of injury and disease names and modifiers, it is difficult to evaluate whether or not each character string is properly coded. is there. However, if the disease name character strings are replaced with structural formulas as in this embodiment, most of them can converge to about several hundred types, so if you evaluate whether it can be adopted for each structural formula, most of the sick name characters It can be determined whether the column can be automatically encoded. Therefore, even when there is no wound name whose wound name matches the wound name master, the user can narrow down the visual confirmation to the minimum.
The evaluation method for each structural formula is as follows. In general, even if a wound name and character name string includes modifiers indicating the wound name and the “position” or “part” of the body, it is considered that these modifiers have little influence on the wound name. For example, even if “right forearm contact dermatitis” is “left lower leg contact dermatitis”, it is still “contact dermatitis”. However, if the included modifiers indicate "etiology" or "condition", these wound names may be expressed in another standard wound name, so they are included in the wound name string. It is not always possible to adopt the names of wounds and diseases. Also, if multiple wound names are detected in the wound name string, it is difficult to determine whether it is just an enumeration of wound names or whether two wound names represent a single meaning. Requires visual confirmation.
Presentation of the coding result evaluation by the structural formula and visual confirmation by the user are performed by the collation result correcting means 12. The collation result correcting means 12 presents the input character string, standardized character string, wound name extracted as a master word / phrase, its code, and structure evaluation obtained by the processing so far to the user. The user determines whether or not the encoding result can be adopted based on these pieces of information, and makes corrections as necessary. These encoding results are stored in the encoding result storage table 13 and output as necessary.
For the correction of the encoded result, the approximate wound name search means 10 and the approximate wound name table 11 are used. The approximate wound name search means 10 performs a fuzzy search in the approximate wound name table 11 for the standardized character string already generated by the notation fluctuation standardization means 4, and extracts a plurality of wound names and codes that approximate the standardized character string. To do. This fuzzy search result is again presented to the user by the collation result modification 12, and the user selects the name of the disease / sickness that seems to be most appropriate from the presented names of the disease / sickness, and the code is encoded. Stored in the save table 13.
According to the prior art, this approximate disease name search operation is performed by the user, and the user searches for an appropriate disease name based on the correspondence table between the disease name and the code, and encodes it by inputting the code. It was. However, this method places a heavy burden on the user's search work, and even if there is a more appropriate name for the disease, the code for another disease name is registered, or an incorrect code is registered due to an input error when entering the code. There was a risk of doing so. In this embodiment, structural evaluation is performed in advance, the minimum necessary input character string that requires visual confirmation is extracted, and the change of the code by the user is performed by selecting the most appropriate wound name from a plurality of approximate wound names. Therefore, it is possible to significantly reduce the burden of coding work by the user and to suppress coding errors due to input mistakes to the maximum.
As described above, according to the present embodiment, it is possible to provide an encoding system that can accurately encode an input character string with fluctuation in notation with less user work load.
Hereinafter, the notation fluctuation standardization means 4 and the notation fluctuation table 5 of this embodiment will be described in detail with reference to FIGS.
FIG. 3 is a diagram showing a specific example of the notation fluctuation table 5 of the present embodiment. The notation fluctuation table 5 stores at least a fluctuation word / phrase 301 that is a variation of conventional notation and kana usage, and a standard word / phrase 302 that is a standard notation corresponding thereto.
The purpose of the notation fluctuation table 5 is to converge the notation fluctuations of various wound names described at the discretion of the doctor to the notation of the master word / phrase stored in the wound name master table 7. Thus, fluctuations in phrases often include variations in notation commonly used in sickness names, such as idiomatic notation (“skin” and “skin-fu”), hiragana / katakana / kanji notation (“biran”, “bilan”, “糜爛” )), Differences in kanji (“neck” and “neck”), pronunciation differences (“virus” and “virus”), presence / absence of long sound symbols (“Qatar” and “catar”), etc. 7 together with a standard word / phrase which is a standard notation registered in 7.
FIG. 4 is a flowchart showing the detailed operation of the notation fluctuation standardization means 4.
First, in step 401, the input character string fetched as the disease name character string data 2 by the control means 14 is fetched into the notation fluctuation standardization means 4.
Next, in step 402, the fluctuation word / phrase 301 in the notation fluctuation table 5 is referenced to search for a word / phrase corresponding to a part of the input character string.
If a corresponding fluctuation word is found, the fluctuation word part is replaced with a word stored in the standard word / phrase 302 of the notation fluctuation table 5 in step 403. When all the fluctuation phrases are converted into standard phrases in steps 402 and 403, the character strings are returned to the control means 14 as standardized character strings in step 404.
The conversion is not necessarily part of the input character string, and the entire input character string is converted into a standard phrase as a fluctuation phrase, or a fluctuation phrase is not detected in the input string and is controlled as a standardized phrase as it is It may be returned to the means 14.
As described above, the notation fluctuation standardization means 4 and the notation fluctuation table 5 standardize the notation fluctuation included in the input character string, and more collation name character strings are used when collating with the wound name master table 6 as the next step. Can be coded.
Compared with the method of storing the disease name character string including notation fluctuation in the user dictionary and the corresponding code as in the conventional technique, in this embodiment, only the notation fluctuation is standardized on the character string, so in the next step It does not depend on the wound name master table 7 to be used.
In other words, when changing the code system used in the disease name master table 7, the user dictionary must change the codes of all registered disease name character strings based on the new code system. It is not dependent on the coding system. Therefore, it is possible to avoid a huge work load that rewrites all the codes in the user dictionary.
Hereinafter, the wound name master collating unit 6 and the wound name master table 7 of the present embodiment will be described in detail with reference to FIGS. 5, 6, 7, 8, and 9.
FIG. 5 is a diagram showing a specific example of the wound name master table 7 in the present embodiment. The wound name / master name table 7 stores at least a master word / phrase 501 such as a wound / illness name and a modifier, a code A 502 and a code B 503 corresponding to the code, and a word / phrase attribute 504 indicating the type of the master word / phrase. There are not necessarily two types of codes to be stored, and one type or three or more types may be stored.
Commonly used codes are the disease name master and modifier master of the receipt computer processing system whose main purpose is to encode the names of wounds and diseases described in the receipt, and many disease names and modifiers are coded. ing. If too many types of codes are used, such as when analyzing disease trends based on the names of wounds and sicknesses, the classification is too fine and 119 items are coded with a large number of wound names. The disease classification (medium classification code) for social insurance charts is often used. As described above, it is preferable that the user can select the code depending on the purpose of use. If a plurality of code systems are stored in the sickness name master table 7, an appropriate code system can be selected according to the user's purpose of use. .
FIG. 6 is a table showing an example of the phrase attribute 504 stored in the wound name master table 7. In the present embodiment, phrase attributes are associated with all master phrases stored in the wound name master table 7. However, since all words that can be freely described as a disease name character string cannot be registered as a master word, a character string that is not registered as a master word is treated as having an unknown word attribute.
FIG. 7 is a flowchart showing the detailed operation of the wound name master collating means 6.
First, in step 701, the standardized character string is input as an input character string by the control means 14 to the wound name master matching means 6.
In step 702, the master word / phrase 501 in the injury / patient name master table 7 is referenced to search for a corresponding word / phrase as a part of the input character string.
If a corresponding word is detected in step 703, the master word 501, code A 502, code B 503, and word attribute 504 are stored in the matching table 15 in step 704. When all corresponding master phrases are stored in the collation table 15, only necessary master phrases are selected in step 705, and the remaining master phrases are deleted from the collation table 15. This is because, for example, there are three types of master phrases that are detected in the phrase “forearm”: “fore”, “arm”, and “forearm”, but only “forearm” is required, and other master phrases are unnecessary. .
In step 706, the master phrase finally selected is compared with the input character string that is the basis thereof, and consecutive partial character strings in which the master phrase is not detected in the input character string are collated as the phrase attribute “unknown”. Store in table 15. At this time, nothing is stored in the codes A and B.
FIG. 8 is a table for explaining rules for selecting only a master word / phrase necessary in step 705 of FIG. When a plurality of master phrases are detected by the master collation, it is checked which one of the master phrases A and B should be selected from the positions and the number of characters of the detected master phrases A and B in the input character string. .
If the relationship between the master word / phrase A and the master word / phrase B is included in the other word / phrase as in 1) to 3) in FIG. 8, the longer master word / phrase is selected. For example, “forearm” is selected for “forearm” and “front”.
If the master phrase A and the master phrase B do not overlap at all in the input character string as shown in 4) of FIG. 8, both are independent master phrases and are selected.
As shown in 6) to 8) of FIG. 8, when the master phrase A and the master phrase B are shifted in the front and back and are arranged in the input character string, the master phrase A or the master phrase B having the larger number of characters is selected. However, only when the number of characters is the same, the one shifted backward as in 6) is selected.
FIG. 9 is a diagram illustrating the state of the collation table 15 before and after the necessary master word / phrase is selected in step 705 of FIG. 7 and the unknown character string is registered in step 706.
In FIG. 9, as an example, it is assumed that a wound name character string “possibility of right forearm contact dermatitis” is input to the wound name master verification unit 6. As a result of the master collation, five master words / phrases “right”, “front”, “forearm”, “arm”, and “contact dermatitis” are detected as master words from the wound name master table. As shown in the table above 9, the collation table 15 stores the respective master words / phrases, code A, code B, and word / phrase attributes.
After that, in step 705 of FIG. 7, only the necessary master word / phrase is selected according to the selection rule shown in FIG. 8, so that only “right”, “forearm”, and “contact dermatitis” remain. Next, as a result of the comparison with the input character string in step 706, it can be seen that the character string “possibility” was not detected as the master word / phrase. As a result of registration in the collation table 15 as a master phrase, the result is as shown in the lower table of FIG.
As described above, the code of each master word / phrase included in the input character string of injury / illness name is acquired by the injury / patient name master collating means 6 and the injury / illness name master table 7. In addition, by acquiring the phrase attribute, the character string structure analyzing means 8 which is the next step can perform the structure analysis, and the encoding result can be evaluated.
Hereinafter, the character string structure analyzing means 8 and the structural formula table 9 of this embodiment will be described in detail with reference to FIGS.
FIG. 10 is a diagram showing a specific example of the structural formula table 9 in the present embodiment. The structural formula table 9 stores at least a structural formula 1001 that is a combination of phrase attributes and a structural evaluation 1002 that indicates whether or not a wound name string having the structural formula can be adopted by the detected wound name code. Has been. In addition, if a comment 1003 which is an explanation of why the structural formula is accepted or not is stored, the comment 1003 is also presented to the user at the time of visual confirmation by the user, and a supplement is made to promote an accurate determination. Can explain.
The structure evaluation 1002 does not necessarily have to be a binary value indicating whether or not it is accepted, and as shown in FIG. Thus, the user refers to the structural evaluation 1002 at the time of visual confirmation, and when it is desired to strictly code, all other than “excellent” is visually confirmed, and when strictness is not required, only “impossible” is visually confirmed. It is possible to use different coding accuracy depending on the situation.
FIG. 11 is a flowchart showing the detailed operation of the character string structure analyzing means 8.
First, in step 1101, the master words / phrases detected in the disease name string in the collation table 15 are rearranged in the order of appearance in the disease name string, and the word / phrase attributes corresponding to each master word / phrase are rearranged. Arrange them in order and connect them to get the structural formula. For example, the structural formula of “left forearm contact dermatitis” is “positional site injury”.
In step 1102, the character string structure analyzing means 8 refers to the structural formula 1001 in the structural formula table 9 and searches for the same structural formula as the structural formula obtained in step 1101.
If the same structural formula is detected in the structural formula table 9 in step 1103, the structural evaluation 1002 and the comment 1003 of the corresponding structural formula are returned to the control means 14 in step 1104.
If the same structural formula is not detected in step 1103, the structural evaluation is returned to the control means 14 as “impossible” and the comment “structural formula is not registered”.
As described above, by using the structural formula that is an array of the phrase attributes of the master phrase included in the input disease name character string by the character string structure analyzing unit 8 and the structural formula table 9, the disease name included in the disease name character string. It is possible to evaluate the structure of whether or not the above code can be used as the code of the disease name character string. Therefore, even if the disease name string does not completely match the master phrase, the user can determine whether it is necessary to visually check the disease name string and the encoding result, based on the structure evaluation, Visual confirmation work can be reduced to the minimum necessary.
It is possible to automatically perform a series of processes for standardizing the fluctuation of the notation of the character string of the disease name input as described above, obtaining a code by master verification, and evaluating a coded result by structure evaluation. Therefore, even if a large number of disease name strings are entered, the user browses the display screen after the automatic processing of all the disease name strings, and visually checks only the disease name character strings that cannot be automatically encoded by the structure evaluation, Other codes can be rewritten as needed.
The collation result correcting unit 12 and the encoded result storage table 13 of this embodiment will be described in detail below with reference to FIGS.
FIG. 12 is a diagram showing a specific example of the encoding result storage table 13 in the present embodiment.
The disease name character strings inputted one after another are coded through each process and accumulated in the coded result storage table 13. The items stored in the encoding result storage table 13 are an input character string 1201 that is an inputted character string as it is, a standardized character string 1202 that is standardized by the notation fluctuation standardization means 4, and a disease name master collating means 6 Among the extracted master phrases, the adopted wound name 1203 whose phrase attribute is “wound and sickness”, the code A1204 which is a code of the adopted wound and sick name 1203, and the code B1205 which is a code according to another code system of the adopted wound and sickness name are continued. The next code A column 1206 is an enumeration of codes indicating the disease name and various modifiers arranged in the order of appearance of the codes of all the master words extracted by the disease name master collating means 6.
Subsequently, the structural formula 1207 obtained by the character string structure analyzing means 8, the structural evaluation 1208, and the comment 1209 are also stored.
In FIG. 12, in order to improve the legibility of the structural formula 1207, a phrase with a phrase attribute “unknown”, a phrase with “parentheses”, and “connection” is displayed using the original character string. That is, in the input character string “asthma-like bronchitis” of number 4, “asthma” and “bronchitis” are detected as master words, but the middle “like” is a word that is “unknown”. Then, “Wound and sickness [like] injury and sickness” is indicated. Similarly, the parenthesis of the parenthesis of “tonsillitis (chronic)” of the number 7 is “parenthesis”, but here is “injury (progress)”.
FIG. 13 is a specific example of a screen displayed on the display unit 16 by the collation result correcting unit 12.
This screen is for presenting the encoded result stored in the encoded result storage table 13 and its evaluation to the user and prompting the user to determine whether or not correction by the user is necessary.
A list 1301 is a list of input character strings accumulated in the encoding result storage table 13 and main items necessary for the user to determine the necessity of correction.
A button 1302 and a button 1303 are used to switch the content displayed in the list 1301 and can select whether to display all the coding results or only the coding results that are visually confirmed based on the structural evaluation. At this time, the structure evaluation for displaying the coding result is set in advance by the user. For example, it is possible to freely increase / decrease the amount of coding correction work by changing the setting so that structural evaluation is “impossible” for some users and “possible” and “impossible” for other users. Therefore, the user can freely control the amount of the name of the visually confirmed disease and sickness according to the number of input character strings, the required encoding accuracy, and the amount of work that can be spent.
Text boxes 1304 to 1312 display data accumulated in the encoding result storage table 13 relating to the numbers selected in the list 1301. The user can make a fine determination based on the information.
Button 1313 causes this input string selected from the list to flow again through a series of steps of automatic encoding, button 1314 adopts this encoding result, and button 1315 does not satisfy this encoding result. This is a button for executing an approximate disease name search when the user decides to code by an approximate disease name search described later. When it is determined that the user adopts this encoded result, it is possible to determine that the coded result is adopted as long as the button 1315 does not perform an approximate wound name search without pressing the button 1314.
In this way, the user can browse only the input character string that seems to be corrected based on the screen displayed by the collation result correcting means 12. Moreover, if it is determined that correction is necessary as a result of careful examination of the coding result, an approximate wound name search screen described later is displayed from this screen, and a more appropriate code can be searched there.
As an example of the approximate character string search in the present embodiment, there is an ambiguous search method using N-grams often used in natural language processing. In general, a character string of length N adjacent in a phrase or sentence is called an N-gram. For example, one gram of the character string “myocardial infarction” is “heart”, “muscle”, “infarction”, and “block”. Gram is “myocardium”, “muscle infarction”, “infarct”. In this way, if the character string is decomposed into grams and the master table is also stored with the master word / phrase decomposed into grams, the larger the number of grams in both, the closer the character strings of both. You can see that Of course, all the grams match if they are the same string. However, even if some letters are missing or the word order is reversed, many grams match. Judge that there is.
The approximate wound name search means 10 and the approximate wound name table 11 in this embodiment will be described in detail below with reference to FIGS.
FIG. 14 is a diagram showing a specific example of the approximate disease name table 11 of the present embodiment.
The approximate word / phrase name table 11 stores master words / phrases stored in the disease / disease name master table 7 in a state of being decomposed into grams, and each word / phrase name master number 1401 manages each master word / phrase in the disease / disease name master table 7. And the gram 1402 stores the gram of the master phrase.
In the example of FIG. 14, 1 gram and 2 gram of “myocardial infarction” of master word / phrase number 1 and “allergic rhinitis” of 2 are stored.
In N-grams, the accuracy of the fuzzy search varies depending on the length N of the character string, but in searching for sickness names, both 1 and 2 are used because of the balance between the search accuracy and the capacity of the table. Just right. Of course, you can use more than 3 grams.
The master phrase for storing the gram in the approximate disease name table 11 does not have to be all the master words stored in the disease name master table 7, and the purpose of the search is to search for the disease name. It is also possible to reduce the number of options and make it easy to select only a certain wound name or a typical wound name. FIG. 15 is a flowchart showing the detailed operation of the approximate disease name search means 10. When the input character string selected by the user on the screen shown in FIG. 13 is instructed to search for an approximate disease name, step 1501 generates 1 and 2 grams of standardized character strings of the disease name character string. At this time, the character string to be gram-decomposed may be the input character string itself, but the search accuracy is improved by using the standardized character string.
In step 1502, the generated gram is collated with a partial character string 1402 stored in the approximate disease name table 11. If there is a matching gram, the number of matched grams is counted for each wound name master number 1401.
Here, it can be said that the more grams match, it can be said that the name of the wound is approximate, but in this embodiment, as the matching rate that quantitatively indicates the degree of approximation of both,
(Number of matched grams) ÷ (number of grams of master phrase)
Is used.
The master word / phrase in which the gram match is found is passed to the collation result correcting means 12, and is displayed to the user together with the master word / phrase, code A, and code B in the descending order of precision using the display means 16 in step 1504.
In step 1505, the user input means 17 is used to select the most appropriate master phrase from these master phrases, and in step 1506, various data of the master phrase is converted into a new encoding result of the input character string. Is stored in the encoding result storage table 13 as follows.
FIG. 16 is a specific example of a screen displayed on the display unit 16 by the collation result correcting unit 12 in step 1504. When the execution of the approximate disease name search is instructed from the screen of FIG. 13, the screen of FIG. 16 is displayed, but the upper half of the input character string list is the same as FIG. The lower half of the screen is a display for searching for an approximate disease name, and an input character string 1501 selected from the list and a standardized character string 1502 are displayed in a text box.
The approximate wound name search result that is the result of the fuzzy search using the gram is displayed in the list on the lower right of the screen, and the candidate wound name 1503 that is the master word / phrase in which the gram match is found, its code A 1504 and code B 1505 , And the precision 1506 (expressed as a percentage).
Of the items displayed in the list, code A 1504 and code B 1505 are not necessarily required, but by comparing with other candidate disease names, it is easier to determine which candidate disease name is appropriate.
Also, by displaying the candidate wound names from the top in the order of high relevance rate, the user can examine the candidate wound names in descending order of coincidence, thereby facilitating selection.
Here as the precision
(Number of matched grams) ÷ (number of grams of master phrase)
Use
(Number of matched grams) ÷ (Number of grams in standardized character string)
The reason for not using is that the order of coincidence between the two is different.
In the former case, for example, when “contact dermatitis” is searched, “contact dermatitis” and “dermatitis” are also displayed at the top of the list with a precision 1 respectively. However, if the latter is used as the precision, “contact dermatitis” is similarly displayed at the top of the list with precision 1; however, “dermatitis” only matches approximately half of the gram. It will be displayed below.
In general, when searching for a wound name is not found, if the exact wound name is not found, the name of the wound that broadly captures the wound name, that is, “dermatitis” for “contact dermatitis”, “acute bronchitis” "It is natural to choose" bronchitis ".
Accordingly, by using the former relevance rate, it is possible to evaluate that the relevance rate is higher for a disease name that broadly captures the name of the disease, and it is possible to present candidate disease names with priority orders suitable for the user's selection work.
When the user finds an appropriate wound name from the list by the approximate wound name search by the automatic coding process, the candidate wound name is selected and the “adopt” button 1507 is pressed, and the result is stored in the coded result storage table. 13 can be stored. If not selected, a “structure analysis” button 1508 is pressed to return to the screen shown in FIG. 13 and a coding result obtained by automatic coding can be adopted.
In order to obtain the code of the disease name string by the approximate disease name search, the user selects the most appropriate one from among tens of thousands of standard injury names as in the conventional technique, and inputs the code by punch input or the like. On the other hand, it is much easier and the effect of preventing erroneous registration due to an input error is high.
In addition, even if an appropriate code cannot be found, it is possible to easily extract the name of the disease and sickness that captures the character string of the disease and disease in a broad sense, so there is little variation in the coding results due to differences in user knowledge and experience, and anyone can It is possible to obtain roughly similar coding results. The consistency of the encoded results is important for statistical analysis because it can prevent differences in results due to differences in interpretation when compared with other disease trend analysis results.
FIG. 17 is a diagram showing a specific example of the coding result storage table 13 after the approximate injury / illness name search in this embodiment.
Here, as a result of an approximate wound name search for “asthmatic bronchitis”, “olfactory dysfunction”, and “tonsillitis (chronic)” in the input character string 1701 whose structural evaluation was “impossible” at the time of FIG. The code has been updated to be more appropriate. Also, the structural evaluation 1708 is changed to “approximate” indicating that it is a result of the approximate injury / illness name search, and this is also displayed in the comment 1709. Since the code A string 1706 and the structural formula 1707 are not generated in the approximate disease name search, the display is blank.
In the approximate wound name search, various character strings of wound names can be encoded. For example, in FIG. 17, "asthmatic bronchitis" (bronchitis with asthma-like symptoms) and "asthmatic bronchitis" (bronchi due to asthma) Injury and disease names that have similar meanings such as “flame”, broader names such as “olfactory dysfunction” and “sensory disturbance”, and word order notation such as “tonsillitis (chronic)” and “chronic tonsillitis” Can be easily coded.
The coding results stored in the coding result storage table 13 can be output to the outside of the computer device 1 arbitrarily, and the user can perform aggregation and statistical processing based on this information and use it for disease trend analysis and the like. it can.
The maintenance screens for various tables in this embodiment will be described below with reference to FIGS. 18 and 19.
FIG. 18 is a diagram showing a specific example of the maintenance screen of the notation fluctuation table 5.
The maintenance screen includes a list that lists words stored in the notation fluctuation table 5, and at least a fluctuation word 1801 and a standard word 1802 are displayed. The conversion count 1803 is an accumulation of the number of times the fluctuation phrase and standard phrase are used by the notation fluctuation standardization means 4.
While viewing this list, the user adds a new fluctuation phrase and standard phrase using the button 1804, deletes an unnecessary fluctuation phrase and standard phrase using the button 1805, changes the contents of the phrase using the button 1806, or changes a button 1807. The maintenance screen can be terminated without editing.
In order to efficiently encode the disease name character string input at any time, it is preferable to always maintain the optimum state of the notation fluctuation table. Therefore, if there is a poor coding result due to a new notation fluctuation, add a new fluctuation phrase, standard phrase, or if the previously registered fluctuation phrase and standard phrase are rarely used thereafter, Deleting it reduces the verification load and enables faster conversion. The number of conversions 1803 is displayed as a guide for grasping the usage status of the fluctuation phrase and the standard phrase.
FIG. 19 is a diagram showing a specific example of the maintenance screen of the structural formula table 9.
On the maintenance screen, there is a list that lists the structural formulas stored in the structural formula table 9, and at least the structural formula 1901 and the structural evaluation 1902 are displayed. A comment 1903 is also displayed to add more detailed information when the structural evaluation is presented to the user.
While viewing this list, the user can add a new structural formula with the button 1904, delete an unnecessary structural formula with the button 1905, change the structural formula or structural evaluation with the button 1906, or do not edit it with the button 1907. The maintenance screen can be terminated.
The structural formula is an index used for evaluating the coding result in this embodiment. When a new structural formula is added or the structural evaluation is not intended by the user, the structural evaluation is adjusted on this maintenance screen. There is a need to do.
The encoding system according to the second embodiment of the encoding method of the present invention will be described below with reference to FIGS.
In the above-described embodiment of the present invention, when the user determines that the result of the automatic coding by the master verification is not appropriate, the user needs to perform an approximate wound name search and convert it to an appropriate code each time. In this case, when the same disease name character string is input, the user is burdened with the same processing.
In the configuration diagram of this embodiment shown in FIG. 20, in addition to the above-described embodiment of FIG. 1, conversion history recording means 18, conversion history search means 19, and conversion history table 20 are provided.
As a result of being automatically encoded and determined to be adopted by the user, or as a result of a search for an approximate disease name by the user, an input disease name character string, standardized character string, code A, stored in the encoded result storage table 13 Code B, code A string, structural formula, structural evaluation, and comment are stored in the conversion history table by the conversion history recording means 18.
In the present embodiment, when a disease name character string is input, the control unit 14 first passes the disease name character string to the conversion history search unit 19, and the conversion history search unit 19 searches the conversion history table for the same input character string. To do.
When the same input character string is detected, the conversion history search means 19 returns the information in the table to the control means 14, and the control means 14 records it in the encoded result storage table 13 via the matching result correction means 12. .
According to the present embodiment, the encoding result once adopted by the user and stored in the encoding result storage table 13 is stored in the conversion history table 20, and when the same injury name character string is input next time, Since the conversion history table 20 is registered in the encoding result storage table 13, various processes in the middle can be omitted. Therefore, particularly when a large number of injury / sickness name character strings are input and many of the same injury / sickness name character strings are included therein, the processing speed can be remarkably improved.
FIG. 21 is a specific example of a screen displayed on the display unit 16 by the collation result correcting unit 12 in the present embodiment.
FIG. 21 is a display screen of an approximate injury / illness name search result similar to FIG. 16 of the above-described embodiment. In FIG. 21, a button 2101 is additionally provided, and the user stores the approximate injury / disorder name search result in the conversion history table 20. It is possible to select whether or not. Therefore, the user can adjust only an arbitrary coding result so as to obtain a similar coding result next time.
The encoding system according to the third embodiment of the encoding method of the present invention will be described below with reference to FIGS.
FIG. 22 is a block diagram of the present embodiment, and in this embodiment, the computer apparatus 1 that performs coding processing on the disease name character string data 2 that is the input of FIG. Connected to the information database 18 to obtain input information. The medical information database 21 is a database for managing medical information such as an electronic medical record or a receipt processing system, and stores various information such as patient information, medical practice information, and accounting information in addition to the names of wounds.
In the present embodiment, since various medical information other than the disease name character string is obtained at the time of encoding, it is possible to refer to such information at the time of visual confirmation and code correction by the user.
FIG. 23 is a specific example of a screen displayed on the display means 18 in the present embodiment. The display unit 17 displays the matching result display screen 2301 described with reference to FIGS. 13 and 15 in the first embodiment, and also displays a medical information display screen 2302 acquired from the medical information database 21. . The medical information display screen 2302 may be text information or image information.
The user confirms the encoding result using the collation result display screen 2301, and refers to the medical treatment information display screen 2302 when making corrections as necessary. It is possible to know and to correct the code with higher accuracy.
The main points of the present invention are summarized as follows.
When an input character string is encoded by master collation, a phrase having a notation fluctuation in the character string is converted into a standard phrase in advance.
Further, in order to convert a phrase having notation fluctuation in the input character string into a standard phrase, the notation fluctuation phrase and the standard phrase are stored in the storage area as a pair.
Furthermore, a display screen for editing a pair of a phrase having a notation fluctuation and a standard phrase stored in the storage area is provided.
When the input character string is encoded by master collation, one or more words / phrases registered in the character string are detected, and an array of the word / phrase attributes is generated. Evaluate whether or not.
Further, the phrase, code and attribute are stored in the storage area.
Further, the phrase attribute and the structure evaluation indicating whether or not encoding is possible are stored in the storage area. Furthermore, it has a display screen for editing the array of phrase attributes and the structure evaluation stored in the storage area.
Furthermore, when the input character string and the encoded result are displayed, only the input character string that needs to be corrected can be displayed as a result of the evaluation.
Further, as a result of the evaluation, if correction is necessary, a screen for performing a fuzzy search from this screen is displayed.
The input character string and the encoded result are stored in the storage area, and the next time the same character string is input, the encoded result is output.
Furthermore, it has a display screen which can select whether to store the input character string and the encoded result in the storage area.
When the input character string and the encoded result are displayed, the form data or image that is the basis of the input character string is displayed together.
[0006]
【The invention's effect】
As described above, the encoding method of the present invention takes time and effort to register all names of wounds and sickness including notation fluctuations of various wound names as in the prior art when coding the names of wounds and sickness. Since there is no need to perform work that is prone to mistakes, the work efficiency and the reliability of code output are greatly improved.
In addition, according to the present invention, as in the prior art, all the names of wounds and sickness that were not detected at the time of matching with the name of the sickness were confirmed by human eyes, and it was necessary to add them to the user dictionary as necessary. By automatically evaluating the encoding result and presenting it to the user, there is an effect that minimum visual confirmation and correction work can be performed according to the user's needs and applications.
In addition, according to the present invention, when coding a disease name that has not been detected at the time of disease name master matching as in the prior art, a code must be directly input by punch input or the like. By selecting the most appropriate approximate disease name, the user records the code, so that the work efficiency and the reliability of code output are greatly improved.
[Brief description of the drawings]
FIG. 1 is a basic configuration diagram showing an embodiment according to the present invention.
FIG. 2 is a process flow diagram illustrating one embodiment according to the present invention.
FIG. 3 is a diagram illustrating an example of a notation fluctuation table.
FIG. 4 is a flowchart of notation fluctuation standardization means.
FIG. 5 is a diagram showing an example of a wound name master table.
FIG. 6 is an explanatory diagram showing types of word attributes.
FIG. 7 is a flowchart of the disease name master collating means.
FIG. 8 is an explanatory diagram showing a selection rule when a plurality of master words / phrases are detected.
FIG. 9 is an explanatory diagram showing the behavior of a collation table.
FIG. 10 is a diagram showing an example of a structural formula table.
FIG. 11 is a flowchart of character string structure analysis means.
FIG. 12 is a diagram showing an example of a coding result storage table after automatic processing.
FIG. 13 is a diagram showing an example of a coding result display screen in the collation result correcting unit after automatic processing.
FIG. 14 is a diagram showing an example of an approximate wound name table.
FIG. 15 is a flowchart of approximate wound name search means.
FIG. 16 is a diagram showing an example of a coding result display screen in the matching result correcting means at the time of approximate disease name search.
FIG. 17 is a diagram showing an example of a coding result storage table after an approximate injury / illness name search.
FIG. 18 is a diagram showing an example of a maintenance screen example of a notation fluctuation table.
FIG. 19 is a diagram showing an example of a maintenance screen of a structural formula table.
FIG. 20 is a basic configuration diagram showing a second embodiment according to the present invention.
FIG. 21 is a diagram showing an example of a coding result display screen in the matching result correcting unit after the approximate injury / illness name search in the second embodiment according to the present invention.
FIG. 22 is a basic configuration diagram showing a third embodiment according to the present invention.
FIG. 23 is a diagram showing an example of a display screen in the third embodiment according to the present invention.
[Explanation of symbols]
1. Computer device
2 ... Injury name string data
3 ... Wound and disease name coding result data
4 ... Notation fluctuation standardization means
5 ... Notation fluctuation table
6 ... Injury and disease name master verification means
7… Injury and disease name master table
8 ... Character string structure analysis means
9 ... Structural formula table
10 .... Approximate wound name search means
11 ... Approximate wound name table
12 ... Verification result correction means
13 ... Encoded result storage table
14 ... Control means
15 ... Collation table
16: Display means
17 ... Input means
18. Conversion history recording means
19 ... Conversion history search means
20 ... Conversion history table
21 ... Medical information database.

Claims (2)

A notation fluctuation standardization means for collating an input character string with a notation fluctuation table, detecting a fluctuation word / phrase in the input character string and converting it to a corresponding standard word / phrase;
The input character string converted into the standard phrase is collated with a wound name master table as a standardized character string, and the wound name and modifier included in the standardized string are detected as a master phrase, and the code of the master phrase is A disease name master collating means to be extracted;
Display means for presenting the input character string, the standardized character string, the wound name extracted as a master word and its code,
The phrase obtained by gram-decomposing the input character string or the standardized character string is compared with the approximate disease name table in which the phrase obtained by gram-decomposing the master word stored in the disease name master table is matched for each master word / phrase. It has an approximate disease name search means for counting the number of grams, using the ratio of the number of matched grams to the number of grams of the master phrase as a precision, and displaying the master phrase based on the precision. Coding system. A program executed on a computer having a recording medium storing a notation fluctuation table and a sickness name master table and a display means ,
The computer collates the input character string with the notation fluctuation table, detects a fluctuation word / phrase in the input character string, converts it to a corresponding standard word / phrase, and uses the input character string converted to the standard word / phrase as a standardized character string. , Collate with the wound name master table, detect the wound name and modifier included in the standardized character string as a master phrase, and extract the code of the master phrase,
The input character string, standardized character string, wound name and disease name extracted as a master word and its code are presented to the display means, and the input character string or the phrase obtained by gram-decomposing the standardized character string is displayed in the disease name and disease name master table. The stored master word / phrase is collated with the approximate disease name table in which the words / phrases are stored, and the number of matched grams is counted for each master word / phrase, and the ratio of the number of matched grams to the number of grams of the master word / phrase is adapted. A program for executing a procedure of presenting the master word / phrase on the display unit based on the relevance ratio, which is used as a ratio.

Images