JP2018169795A - Information processing device, image display method therefor, and analysis processing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technology of analyzing a medical information database including unstructured data.SOLUTION: An information processing device vectorizes text information written in a medical record in a medical region such as a mental disorder region, by calculating the frequency of a word included in the text information by predetermined natural language processing, and associates a writing timing indicating a timing at which the medical record was written, the word, and the frequency with each other.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an information processing apparatus that uses medical data to efficiently search for medical needs in a mental illness area.

  In recent years, projects have been started in various countries to collect and analyze medical data such as hospital information system records and insurance claims, and to help evaluate the safety of pharmaceuticals. In Japan, there is a business that collects medical information data from multiple facilities, builds a database, and analyzes it. For example, based on the analysis result of a database of medical information, a search for a treatment policy that achieves both effective treatment and medical cost control and development of new insurance products are being carried out. In addition, there is a movement in pharmaceutical companies to search for unmet medical needs (medical needs for which there is still no effective treatment method) (new drug development, etc.).

  As a technique for analyzing a database of medical information, for example, there is a technique described in International Publication WO 2010/032343 (Patent Document 1). In this document, “similarities are calculated in consideration of the concept of time series and similar cases are searched. An event composed of data representing a patient's condition or treatment is stored in a database, and For each event, data to be searched is generated in which the events that occurred until the event occurred are arranged in time series, while a search query in which events are arranged in time series is also generated for the search target patient. There is a description that “similar case search system that calculates similarity between query and searched data and displays searched data in order based on similarity” is provided.

  Among medical information, information such as a psychiatric area (mental illness area) is characterized by a description in a medical record. For example, in psychiatry, diagnoses are based on careful examinations, so psychiatrists are not only patient complaints but also patient attitudes and facial expressions as important patient information obtained as a result of the examination. It is written in the medical record in text, and the patient is diagnosed, evaluated for severity, and treated against typical cases.

International Publication WO2010 / 032343

  Many diseases in the medical field are analyzed and diagnosed by indicators based on scientific evidence such as clinical tests, and treatment is performed. However, for example, psychiatric diseases (psychiatric disorders) such as depression and schizophrenia are under development based on scientific evidence, and are based on clinical tests compared to non-psychiatric disorders. Difficult to carry out analysis and diagnosis. For this reason, there are a wide variety of drugs that are targets for the development of new drugs for diseases in the psychiatric field, and new drug development has not progressed very easily.

  In addition, the patient information described in the psychiatric chart includes unstructured data that is a sentence expressing the patient's attitude, facial expression, and the like, so the amount of description is enormous compared to other clinical departments. For example, there is an estimate that 500 minutes are required to read all the records of a patient who has been hospitalized three times for three months. For this reason, the technique disclosed in Patent Document 1, such as various test values, is structured as time-series data, and the technology for storing events composed of this time-series data is not structured such as medical information in the psychiatric field. It is difficult to extract information corresponding to an event from the data.

As described above, for example, a technique for analyzing a database including unstructured data such as medical information in the psychiatric field has attracted attention.
An object of the present invention is to provide a technique for analyzing a database of medical information including unstructured data.

In order to solve the above problems, one of information processing apparatuses according to the present invention is an information processing apparatus that analyzes changes in patient symptoms based on medical records.
The information processing apparatus calculates the frequency of words included in the text information by predetermined natural language processing on the text information described in the medical record in a predetermined medical area, vectorizes the text information, A text vectorization processing unit is provided that creates text vectorization information that associates a description timing indicating a description timing with a word and a frequency.

  ADVANTAGE OF THE INVENTION According to this invention, the technique which analyzes the database of medical information containing unstructured data can be provided. Problems, configurations, and effects other than those described above will be clarified by the following description of embodiments.

It is an example of the flowchart explaining the analysis process of the medical record in the mental disease area | region by information processing apparatus. It is an example of the block diagram of information processing apparatus. It is an example of the prescription information table contained in a clinical database. It is an example of the medical record information table contained in a clinical database. It is an example of a text vectorization table. It is an example of a text structure table. It is an example of a symptom model parameter table. It is an example of a symptom occurrence probability table. It is an example of a waveform pattern dictionary table. It is an example of a prescription pattern information table. It is an example of the flowchart of the administration continuation period calculation process which calculates the continuation period of a chemical | medical agent administration event. It is an example of a display screen of a change in prescription for each patient. It is an example of a display screen of a change in prescription for each patient. It is an example of the flowchart of the symptom model creation process which creates a symptom model. It is an example of the flowchart of the occurrence probability calculation process which calculates the occurrence probability of a symptom from each record of a medical record table. It is an example of a display screen of change in prescription for each patient and occurrence probability of symptoms for each patient. It is an example of the flowchart of the waveform pattern approximation process which produces | generates a waveform with the daily occurrence probability of a certain symptom, and approximates with a waveform pattern. It is an example of a screen display which shows the change of the patient state before and after medicine administration. It is an example of the flowchart of a symptom change display process which visualizes a medicine administration event and a symptom change to a user interface. It is an example of a screen display which shows a medicine administration event and a change of a symptom.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. Note that components having the same function are denoted by the same names and reference numerals in the drawings for describing the embodiments, and repetitive description thereof is omitted.

  In addition, as an example of unstructured data, the case of medical information in the mental illness area will be described. However, the present invention is not limited to this, and medical information including unstructured data is medical information of other clinical departments. May be.

FIG. 1 is an example of a flowchart for explaining medical record analysis processing in a mental illness region by the information processing apparatus.
The information processing apparatus 1 according to the present embodiment includes words included in text information by performing predetermined natural language processing on text information such as a patient's state recorded at the time of medical treatment by a doctor among medical records (medical charts) in a mental disease region. The text information is vectorized by calculating the frequency (S2), and the description timing (description date, description time (description time), description date, etc.) indicating the timing at which the medical record is described is associated with the word and the frequency. Text vectorization information is created (S3), based on a symptom model for each symptom prepared in advance including the word related to the symptom characteristics of the psychiatric disease region and the frequency of the word, and the word and frequency of the text vectorization information. Then, the occurrence probability that is the probability that the symptom is occurring is calculated for each description timing (S4), and the occurrence probability for each symptom and each description timing. Output (S5).
Next, details of the information processing apparatus 1 according to the present embodiment will be described.

FIG. 2 is an example of a configuration diagram of the information processing apparatus. The information processing apparatus 1 includes a control unit 10, an input / output interface (I / F) 12, a storage unit 11, and a communication I / F 13 connected by an internal bus.
The control unit 10 includes a prescription event processing unit 101, a text structuring processing unit 102, a symptom model creation unit 103, a symptom occurrence probability calculation unit 104, a waveform pattern approximation unit 105, and a text vectorization processing unit 106. And the respective units (storage unit 11, input / output I / F 12 and communication I / F 13) in the information processing apparatus 1 are controlled.

The storage unit 11 analyzes the clinical database 14 that records the clinical information of the patient (the prescription information table 20 that stores the patient's prescription information and the medical record table 30 that records the medical record by the doctor), and the analysis result of the medical record table 30. The text vectorization table 40 to be recorded, the text structuring table 50 to record the result of structuring the description contents of the medical record table 30, and the symptom used to determine which symptom the description contents of the medical record table 30 correspond to. A symptom model parameter table 60 that records parameters for each model (symptom model), a waveform pattern dictionary table 70 that records changes in symptoms (changes in patient state) as a waveform, and a prescription that records the results of analysis of the prescription information table 20 This is a storage device for storing the pattern information table 80. Information stored in the clinical database 14 is input to the information processing apparatus 1 by a doctor or the like via the input / output interface (I / F) 12 and / or the communication I / F 13. Details of each table will be described with reference to FIGS.
The input / output I / F 12 is connected to various input / output devices (not shown), and receives various inputs from the user via the input / output device according to control from the control unit 10, and processing results in the control unit 10. Is output. The communication I / F 13 is connected to a wired (or wireless) communication line (not shown), and communicates with an external server or the like according to control from the control unit 10.

The control unit 10 is a control device configured by a processor such as a Central Processing Unit (CPU). The processor operates as each functional unit in the control unit 10 by reading and executing a program stored in the storage unit 11 or the like. Data (including a program) necessary for the processor to execute processing may be stored in advance in the memory (the storage unit 11 or another storage device (non-temporary or temporary storage device) (not shown)). The load may be loaded (installed) from a server or the like via the communication I / F 13 when necessary.
Further, the control unit 10 and the storage unit 11 may be configured by designing with an integrated circuit such as a field-programmable gate array (FPGA) or an application specific integrated circuit (ASIC), for example.

FIG. 3 is an example of a prescription information table included in the clinical database. The prescription information table 20 stores information (data) related to prescription of medicines for each patient. Specifically, patient identification (ID) 21 that identifies each patient, prescription date 22 that indicates the date on which each patient was prescribed, prescription ID 23 that identifies the prescribed prescription, the name of the prescribed drug, etc. Are registered data such as a prescription name 24, a prescribed drug dosage 25, a drug administration unit 26, and the like.
The prescription date 22 may be a prescription time (prescription time) or a timing at which another prescription such as a prescription date is performed.

FIG. 4 is an example of a medical record table included in the clinical database. The medical record table 30 stores information (data) related to records during medical treatment by a doctor. Specifically, a medical record text that is text information recorded by a doctor at the time of medical treatment regarding a patient ID 31, a description date 32 indicating a date on which medical record information is described, a patient's condition and objective judgment, evaluation, and treatment policy 33 is data to be registered.
Note that the description date 32 may be a description time (description time) or other description timing such as a description date.

  As shown in FIG. 4, the medical record text 33 is obtained from S (subjective): subjective information (information obtained from the patient's story), O (objective): objective information (physical examination, examination) Information), A (assessment): Evaluation, diagnosis (what can be considered from O and S), P (plan): SOAP format described in terms of plan (treatment policy, content, life guidance) is common.

In psychiatric reports, S often contains a conversation between a patient and a healthcare professional. The doctor makes an objective judgment based on the information obtained from the conversation with the patient, and records O, A, and P.
For simplification, only patient ID 31 “P005” is shown in FIG.

FIG. 10 is an example of a prescription pattern information table. The prescription pattern information table 80 stores information obtained by patterning prescription information based on the prescription information table 20. Specifically, it is data in which a patient ID 81, a prescription ID 82, a prescription start date 83, and a prescription end date 84 are registered. By grouping with the prescription ID 82 and the start date 83, the prescription information is patterned to make it easy to tabulate.
Note that the start date 83 may be a start time (start time) or a timing at which another prescription such as a start date and time is started. The same applies to the end date 84.
The prescription pattern information table 80 is created by the prescription event processing unit 101 by the process of FIG.

  FIG. 5 is an example of a text vectorization table. The text vectorization table 40 is created by the text vectorization processing unit 106 based on the medical record table 30 by known morphological analysis. In this table 40, the contents described in S in the medical record text 33 of the medical record table 30 in particular are included in the text after being subjected to natural language processing (for example, morphological analysis, word separation, N-gram, etc.) and word separation. Stored is a vectorized (numbered sequence) of text words using the frequency (appearance frequency) of the generated words as a value. Specifically, a patient ID 41, a description date 42, and a word frequency 43 included in the medical record text 33 are stored.

  FIG. 6 is an example of a text structured table. The text structuring table 50 is created based on the medical record table 30. In this table 50, information obtained by structuring each data stored in the medical record table 30 is stored. The data registered in the text structuring table 50 corresponds to the patient ID 51, the description date 52, the section title 53 indicating the section described in the medical record text 33, and the contents described in the section. And the presence / absence 55 of the symptom.

  In this embodiment, the section title 53 is a case where the medical record text 33 is described in the SOAP format, and SOAP is regarded as a section, and “S”, “O”, “A”, and “P” indicate which of the medical record text 33 Indicates what was in the section.

  The text structuring table 50 is created by the text structuring processing unit 102. For example, the text structuring processing unit 102 stores a symptom name in the symptom 54 using a pre-prepared symptom dictionary (not shown), and uses a pre-prepared symptom appearance determination dictionary (not shown) for medical treatment. Whether or not “symptom” is described from the recorded text 33, that is, “presence / absence” of symptom 54 is determined and stored in presence / absence 55. If a symptom ID is given to various dictionaries, a symptom ID may be input to the symptom 54. The presence / absence 55 stores, for example, 1 when there is a symptom and 0 when there is no symptom.

  FIG. 7 is an example of a symptom model parameter table. The symptom model parameter table 60 stores information indicating model parameters for each symptom since the parameters of the symptom model differ for each symptom. Specifically, it is data in which a symptom 61 for which a symptom model is to be generated and a file path 62 for identifying a storage location of a symptom model file created when the symptom model of the symptom 61 is created are registered. The parameter of the symptom model is a word that suggests a symptom and its frequency. The symptom model parameter table 60 is created by the symptom model creation unit 103 by the process of FIG.

  FIG. 8 is an example of a symptom occurrence probability table. In the symptom occurrence probability table 65, the probability of occurrence of a symptom (occurrence probability) is stored for each symptom. Specifically, the patient ID 66, the description date 67, the symptom 68, the occurrence probability 69 of the symptom, and the like are registered. The symptom occurrence probability table 65 is created by the symptom occurrence probability calculation unit 104 by the process of FIG.

  In addition, when patient ID21,31,41,66 is the same value, it has shown that it is the same patient. The same applies to the description dates 32, 42, 67, prescription IDs 23, 82, and symptoms 54, 61, 68.

  FIG. 9 is an explanatory diagram showing an example of a waveform pattern dictionary table. The waveform pattern dictionary table 70 stores information in which a waveform that is a daily symptom change (patient state change) is converted into a dictionary as a pattern. Specifically, the waveform ID 71 for identifying the waveform, the function 72 for approximating the waveform, the parameter value polarity 73 indicating the polarity of the parameter value included in the function, the improvement tendency 74 indicating the change tendency of the symptom, and the like are registered. is there. In the case of FIG. 9, the parameter value polarity 73 indicates the polarity of the parameter value “a” included in the function 72. The waveform pattern dictionary table 70 is created in advance.

Next, detailed operation of each functional unit of the control unit 10 will be described.
FIG. 11 is an example of a flowchart of an administration duration calculation process for calculating the duration of a drug administration event. This process is executed by the prescription event processing unit 101 for each patient ID.
First, the prescription event processing unit 101 reads from the storage unit 11 a list of drugs (initially administered drug list) (not shown) to be administered (prescription) in the early stage of a disease prepared in advance (S111). The prescription event processing unit 101 searches based on the initial administration drug list and the prescription information table 20 for the drug with the earliest administration date (prescription date) among drugs administered in the early stage of the disease (S112).

Specifically, the prescription event processing unit 101 reads the prescription information table 20 from the storage unit 11, and uses the prescription ID of the drug to be administered at the early stage of the disease described in the initial administration drug list as a search key, and the prescription information table 20 The prescription ID 23 is searched, and the one with the earliest prescription date is searched (S112). For example, in the case of the patient ID “005” in FIG. 3 and the drugs “N05AA01” and “N05AX08” described in the initial administration drug list, the prescription event processing unit 101 searches for “N05AA01” or “N05AX08”. The prescription ID 23 is searched for as a key, and the earliest prescription date of the prescription dates 22 of the hit record is further searched (S112).
The prescription event processing unit 101 then registers the prescription ID 23 and the prescription date 22 of the earliest prescription date searched for in the prescription ID 82 and the start date 83 of the record in which the patient ID 81 of the prescription pattern information table 80 is the same value as the patient ID 21. (S113).

  Next, the prescription event processing unit 101 searches for the next record after the administration start date (S114). Specifically, the prescription event processing unit 101 searches for the next record after the prescription date (administration start date) of the start date 83 registered in S113 among the prescription dates 22 (S114). When there is another drug administration, that is, when the prescription ID 23 of the record hit in S114 is different from the drug used as the search key in S112 (YES in S115), the prescription event processing unit 101 determines the administration period of the drug Is registered in the end date 84 of the prescription pattern information table 80 with the day before the prescription date 22 of the next medicine (prescription ID 23 of the record hit in S114) as the end date (S116).

When the same medicine, that is, the prescription ID 23 of the record hit in S114 and the medicine used as the search key in S112 are the same (NO in S115), the prescription event processing unit 101 sets the prescription date 22 of the record hit in S114. Based on the administration start date, the administration interval of the same medicine is calculated. If it is equal to or less than a predetermined value (YES in S117), the process returns to S114 and further searches for the next record.
If the administration period of the same drug is equal to or greater than the predetermined value (NO in S117), the prescription event processing unit 101 determines that the administration period of the drug has ended, and sets the prescription date 22 of the record hit in S114 as the end date. 84 is registered (S116).

  Next, the prescription event processing unit 101 searches the next record in the prescription information table 20, and if there is a value in the next medicine administration, that is, the prescription ID 23 (YES in S118), the prescription ID 23 and the prescription of this record The date 22 is registered in the prescription ID 82 and the start date 83 of the next record in the prescription pattern information table 80 (S119), and the process returns to S114 to further search for the next record. If the next medicine is administered, that is, if the prescription ID 23 has no value (NO in S118), the prescription event processing unit 101 ends the process.

Also, the end date may be input assuming that the administration period has been switched when the administration of the drug is increased or decreased. Specifically, in S117, the prescription event processing unit 101 calculates a change in dosage based on the dosage 25 of the record hit in S114 and the dosage 25 of the record of the administration start date, and calculates the administration period. Switching may be determined. This process is performed when obtaining a combination of a period during which there is no change in the drug and the drug being administered. For example, a prescription combination pattern can be obtained by grouping by start date and patient ID.
The administration period of the drug can be obtained by the administration duration calculation process as described above.

  FIG. 12 is an example of a display screen for changes in prescription for each patient. The screen of FIG. 12 displays a change in prescription (change of administered drug) for the patient by the period 93 in which the prescription of the patient ID 91, prescription ID 92, and prescription ID 92 is performed. This display screen is generated by the prescription event processing unit 101 based on the prescription pattern information table 80 generated by the processing of FIG. 11, and is output to a display device that is an input / output device.

  12A and 12B are display screens for patient IDs 91 “P005” and “P006”, respectively. In FIG. 12A, the prescription ID 92 has changed from “N05AA01” to “N05AX08”, and has changed to “N05AH03” and “N05AH04”. In FIG. 12B, the prescription ID 92 “N05AX08” is changed to “N05AH03” and “N05AH04”, and is changed to “N05AH02”.

  FIG. 13 is an example of a flowchart of a symptom model creation process for creating a symptom model. This process is executed by the symptom model creation unit 103.

  The symptom model creation unit 103 selects one item from the symptom dictionary as a symptom for which a symptom model is to be created (S121). The symptom model creation unit 103 searches the symptom 54 of the text structuring table 50 using the selected symptom as a search key, and acquires a corresponding record (S122). As a result, the case selected in S121 is divided into the case with the symptom (record hit in the search) and the case without the symptom (record not hit in the search).

  The symptom model creation unit 103 searches the description date 42 of the text vectorization table 40 using the value of the description date 52 of each record of the acquired case of the symptom as a search key, and acquires the frequency 43 of the corresponding record (S123). ). The symptom model creation unit 103 creates a symptom model by logistic regression analysis of the frequency 43 corresponding to a symptom case (S124). In S124, other analysis methods may be used. The symptom model creation unit 103 saves the parameter values of the created model in a file, and stores the file path and file name in the symptom model parameter table (S125).

  The symptom model creation unit 103 determines whether all items in the symptom dictionary have been selected (S126). If there are other items (NO in S126), the process returns to S124 to select the next item (S121). . If there are no other items (YES in S126), the process ends.

  FIG. 14 is an example of a flowchart of an occurrence probability calculation process for calculating an occurrence probability of a symptom from each record of the medical record table. This process is executed by the symptom occurrence probability calculation unit 104 for each patient. This process is executed at a predetermined timing (such as every day). Hereinafter, the patient ID “P005” will be described as an example.

  The symptom occurrence probability calculation unit 104 selects one symptom from the symptom dictionary as a symptom occurrence probability to be calculated (S131). The symptom occurrence probability calculation unit 104 searches the symptom 61 in the symptom model parameter table 60 using the selected symptom as a search key, acquires the file path 62 of the corresponding record, and selects the access by following the acquired file path. A symptom model file corresponding to the symptom is obtained (S132).

  Then, the symptom occurrence probability calculation unit 104 searches the patient ID 41 of the text vectorization table 40 using the patient ID “P005” as a search key, and obtains the data of the patient (the value of the description date 42 and the frequency 43) (S133). ). The symptom occurrence probability calculation unit 104 reads the symptom model from the file obtained in S132, and calculates the symptom occurrence probability based on the data of the patient (S134). The patient ID, description date, symptom, and symptom occurrence probability are stored in each item of the symptom occurrence probability table (S135).

The symptom occurrence probability calculation unit 104 determines whether all items in the symptom dictionary have been selected (S136). If there are other items (NO in S136), the process returns to S124 to select the next item (S131). ). If there are no other items (YES in S136), the process is terminated.
FIG. 15 is a display screen example of prescription change for each patient and occurrence probability of symptoms for each patient. The screen of FIG. 15 displays changes in symptoms (changes in patient state) according to a period 97 in which prescription is performed for the patient ID 95, symptom 96, and patient ID 95. This display screen is generated by the symptom occurrence probability calculation unit 104 based on the symptom occurrence probability table 65 generated by the processing of FIG. 14, and is output to a display device which is an input / output device.
In FIG. 15, for the patient with patient ID 95 “P005”, the occurrence probability calculated by the symptom model of “excitement” and “delusion” for the symptom 96 is displayed in the period 97. This occurrence probability is calculated on a daily basis based on daily medical records. In addition, you may display the prescription change of FIG. 12 of the same patient ID with the occurrence probability of FIG.

  FIG. 16 is an example of a flowchart of a waveform pattern approximation process in which a waveform is generated based on the daily occurrence probability of a certain symptom and approximated by a waveform pattern. As a result of this processing, FIG. 17 shows a display example in which a waveform representing a change in a certain symptom is approximated by a function and output to a display device. This process is executed for each patient by the waveform pattern approximation unit 105. Hereinafter, the patient ID “P005” and the symptom “delusion” will be described as an example.

  The waveform pattern approximating unit 105 acquires a waveform based on the occurrence probability of the symptom calculated daily by the occurrence probability calculation process of FIG. 14 (S151). The waveform pattern approximating unit 105 refers to the prescription pattern information table 80 that stores the duration of drug administration, and divides the acquired waveform before and after drug administration (prescription switching) (S152). Specifically, the waveform pattern approximating unit 105 searches the patient ID 81 in the prescription pattern information table 80 using the patient ID “P005” as a search key, obtains the patient data (value of the start date 83), and Each date for which the probability is calculated is compared with the value of the start date 83, the date before and after the drug administration is specified, and the waveform is divided before and after the drug administration (S152).

The waveform pattern approximating unit 105 refers to the function 72 and the parameter value polarity 73 of the waveform pattern dictionary table 70 for each before and after the waveform prescription switching acquired in S151, that is, the fitting by the function, that is, acquired in S151. The waveform is approximated by the function 72 indicating the waveform closest to the waveform before or after prescription switching (S153). Then, the waveform pattern approximation unit 105 refers to the waveform pattern dictionary table 70, acquires an improvement tendency 74 corresponding to each of the approximated functions 72 (S154), and includes the function approximated to the waveform acquired in S151. It outputs so that it may display on a display apparatus as GUI like FIG. 17 which shows the change of the symptom before and behind medicine administration (S155).
The acquired improvement trend 74 may also be output to be displayed on the display device.

  FIG. 17 is a screen display example showing changes in symptoms before and after drug administration. In this display example, the waveform 161 acquired in S151, the functions 162A and 162B approximated in S153, and the date (medication change) 163 divided by S152 are displayed.

  In the case of medical charts written in areas other than the mental illness area, the keyword is registered in the dictionary, and it is determined whether or not a condition related to the keyword is occurring depending on the presence of the registered keyword or whether the keyword is affirmed or denied. can do. On the other hand, in the mental illness area, the description of the medical chart is mainly a conversation sentence, and it is difficult to grasp the presence or absence of a symptom by a keyword included in the conversation. For example, even if it is a conversational sentence indicating a delusional state, it is difficult to define a keyword in the dictionary because the contents shown for each patient vary greatly.

  In addition, since the doctor's judgment at the time is recorded together with the recording of the conversation sentence in the chart of the mental illness area, the judgment of the doctor is helpful as well as the conversation. However, doctors' judgments are not always described when all symptoms occur. Therefore, the occurrence probability of the symptom is calculated from the relationship between the description with the judgment of the doctor and the conversation, and the symptom is numerically expressed as the occurrence probability even when there is no judgment of the doctor. Before and after a prescription event, even if the symptoms are improving, the symptoms do not always decrease, but they get better and worse. Therefore, in order to determine whether or not there is an improvement trend by analyzing the change in symptoms with a trend over a certain period, the time-series changes in numerical symptoms are approximated by a function. Thereby, the tendency of a symptom can be caught.

  FIG. 18 is an example of a flowchart of a symptom change display process for visualizing a drug administration event and a change in symptom on the user interface. As a result of this processing, FIG. 19 shows a display example in which drug administration events and changes in symptoms are output to the display device. This process is executed by the prescription event processing unit 101.

  The prescription event processing unit 101 receives an input of a drug name noted by the user from the user via the input / output I / F 12 or the communication I / F 13 (S141), and to which drug administration event including the drug name is displayed. Is received (S142). For example, it is assumed that up to N events are displayed. In addition, you may receive prescription ID in S141.

  The prescription event processing unit 101 searches the prescription ID 82 in the prescription pattern information table 80 using the prescription ID corresponding to the drug name received in S141 as a search key, and extracts the start date 83 of the corresponding record (S143). The prescription event processing unit 101 extracts, from each record of the prescription pattern information table 80, from the start date of the prescription ID to N event records (S144).

The prescription event processing unit 101 outputs a transition pattern (sequence pattern) to another event associated with the first event extracted in S143 to the display device (S145), and selects a sequence pattern of an event that the user is interested in. Accept (S146). The prescription event processing unit 101 collects, as patient episode data (patient history information), events from the prescription pattern information table 80 that correspond to the sequence pattern selected in S146 (for example, transition from prescription N05AX08 to prescription N05AH02) ( In step S147, the data is aggregated for each episode data and output as a GUI as shown in FIG. 19 for display on the display device (S148).
Episodic data may be displayed for each patient. In this case, changes in the prescription (FIG. 12) and prescription changes and symptom occurrence probabilities (FIG. 15) are applied to the patient corresponding to the series pattern selected in S146. ) May be displayed.

  FIG. 19 is a screen display example showing a drug administration event and changes in symptoms. In this display example, the medicine corresponding to the prescription ID “N05AX08” is input to the information processing apparatus 1 as the name of the medicine to which the user pays attention, and the change in the medicine administration event and the symptom when the display target is one event is visualized. . The case where the drug was switched from the prescription ID “N05AX08” to “N05AH03” and “N05AH04” and the case where the drug was switched from “N05AX08” to “N05AH02” are shown. In each case, 120 episode data and 100 patterns Indicates that there is episode data. In addition, before and after the event switching, that is, how the symptoms changed in each case as a result of the prescription change (prescription change history) (symptom change degree) "improvement", "no change" “Deterioration”.

  “Improvement”, “no change”, and “deterioration” may be determined by the information processing apparatus 1 or the user based on test values obtained by various tests. For “improvement”, “no change”, and “deterioration”, n months before and after the event switching may be designated to display the change result of the patient state. As an example, the occurrence probability of symptoms before event switching and the occurrence probability after event switching are averaged. “Deteriorated” is displayed when it is deteriorating, and “Improved” is displayed when the average occurrence probability is improved before and after event switching. As described above, it is possible to analyze a database of medical information in a mental illness area where there is no clear objective index for measuring the severity or therapeutic effect. That is, in the psychiatric field, the symptom can be quantified based on the medical record description. This makes it possible for doctors to quantitatively understand changes in symptoms (patient status) before and after drug administration, how the effect changes depending on the type of drug to be administered, patient background, Groups of patients with similar changes can be collected and analyzed. Furthermore, by extracting patient groups that are effective due to prescriptions and patients who are not effective, and analyzing the background factors of patients who are not effective, it is possible to narrow down the target of drug development in the psychiatric disorder area. , Which can lead to the acceleration of new drug development.

  In addition, this invention is not limited to an above-described Example, Various modifications are included. For example, the above-described embodiments have been described in detail for easy understanding of the present invention, and are not necessarily limited to those having all the configurations described. Further, a part of the configuration of one embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of one embodiment. Further, it is possible to add, delete, and replace other configurations for a part of the configuration of each embodiment. Each of the above-described configurations, functions, processing units, processing means, and the like may be realized by hardware by designing a part or all of them with, for example, an integrated circuit. Each of the above-described configurations, functions, and the like may be realized by software by interpreting and executing a program that realizes each function by the processor. Information such as programs, tables, and files for realizing each function can be stored in a recording device such as a memory, a hard disk, or an SSD (Solid State Drive), or a recording medium such as an IC card, an SD card, or a DVD.

10: control unit, 101: prescription event processing unit, 102: text structuring processing unit, 103: symptom model creation unit, 104: symptom occurrence probability calculation unit, 105: waveform pattern approximation unit, 20: prescription information table, 30: Medical record table, 40: text vectorization table, 50: text structuring table, 60: symptom model parameter table, 70: waveform pattern dictionary table, prescription pattern information table

Claims (15)

An information processing apparatus for analyzing changes in patient symptoms,
The frequency of words included in the text information is calculated by the predetermined natural language processing on the text information described in the medical record in a predetermined medical area, the text information is vectorized, and the medical record is described A text vectorization processing unit for creating text vectorization information that associates the description timing indicating the timing with the word and the frequency;
An information processing apparatus comprising: The information processing apparatus according to claim 1,
Based on the symptom model for each symptom prepared in advance including the word related to the symptom characteristic of the medical treatment area and the frequency of the word, and the word and the frequency of the text vectorization information, the symptom is surely occurring. A symptom occurrence probability calculation unit for calculating the probability of occurrence for each described timing,
An information processing apparatus further comprising: An information processing apparatus according to claim 2,
For any symptom, based on the occurrence probability for each described timing calculated by the symptom occurrence probability calculation unit, obtain a waveform indicating the change in the occurrence probability, which is a change in the symptom, and approximate the waveform A waveform pattern approximation unit that performs approximation by the function on the acquired waveform with reference to waveform pattern dictionary information that has been previously dictionaryd as a function pattern,
An information processing apparatus further comprising: The information processing apparatus according to claim 3,
About prescription information prepared in advance concerning the prescription including prescription timing indicating a timing date when prescription of medicine to the patient is performed and prescription identification information indicating the prescription performed, the prescription among a plurality of the prescription timings A prescription event processing unit that calculates a start timing at which the prescription is started based on a prescription timing group having the same identification information and creates prescription pattern information that associates the prescription identification information with the start timing,
The waveform pattern approximation unit is
Compare the start timing of the prescription pattern information and the description timing at which the occurrence probability indicated by the waveform is calculated, identify the timing before and after the prescription of the medicine, delimit the waveform before and after the prescription, respectively before and after the prescription To make the approximation,
Information processing device. The information processing apparatus according to claim 4,
The waveform pattern approximation unit is
Output the waveform and the function approximated to the waveform before and after the prescription,
Information processing device. The information processing apparatus according to claim 1,
The prescription identification among a plurality of the prescription timings for prescription information prepared in advance relating to the prescription including prescription timing indicating the timing at which prescription of medicine to the patient is performed and prescription identification information indicating the prescription performed. A prescription event processing unit for calculating a start timing at which the prescription is started based on a prescription timing group having the same information, and creating prescription pattern information that associates the prescription identification information with the start timing,
The prescription event processing unit
When an input specifying an arbitrary medicine is received, the prescription identification information indicating the prescription of the arbitrary medicine and the prescription identification indicating the prescription of the medicine prescribed next to the arbitrary medicine are referred to the prescription pattern information. Extracting at least one piece of information and outputting it as a change history of the prescription of the medicine;
Information processing device. The information processing apparatus according to claim 6,
Based on the symptom model for each symptom prepared in advance including the word related to the symptom characteristic of the medical treatment area and the frequency of the word, and the word and the frequency of the text vectorization information, the symptom is surely occurring. A symptom occurrence probability calculation unit for calculating the probability of occurrence at each described timing,
The prescription event processing unit
Along with the change history of the prescription, output the degree of change of the symptom based on the occurrence probability calculated by the symptom occurrence probability calculation unit,
Information processing device. The information processing apparatus according to claim 1,
The prescription identification among a plurality of the prescription timings for prescription information prepared in advance relating to the prescription including prescription timing indicating the timing at which prescription of medicine to the patient is performed and prescription identification information indicating the prescription performed. A prescription event processing unit for calculating a start timing at which the prescription is started based on a prescription timing group having the same information, and creating prescription pattern information that associates the prescription identification information with the start timing,
The prescription event processing unit
According to the prescription identification information of the created prescription pattern information and the prescription timing, output data for displaying a screen showing a change of a prescribed medicine,
Information processing device. The information processing apparatus according to claim 1,
A text structuring processor that calculates the presence / absence of symptoms based on the text information, and creates text structured information for structuring the text information by associating the description timing with the symptoms and the presence / absence of the symptoms; ,
A symptom model creating unit that prepares the symptom model in advance based on the text structuring information and the text vectorization information;
The symptom model creation unit
Selecting a symptom for which the symptom model is to be created, searching the text structured information to obtain the description timing corresponding to the selected symptom, searching the text vectorization information, and the acquired description timing; Acquiring the corresponding frequency, and analyzing the acquired frequency by a predetermined analysis method to create the symptom model,
Information processing device.   A screen display device that displays a screen having a symptom of a patient in a predetermined medical treatment area and an occurrence probability that is a probability of occurrence of the symptom calculated by a predetermined calculation process.   A screen display method for displaying a screen having a symptom of a patient in a psychiatric disorder region, an occurrence probability that is a probability of occurrence of the symptom calculated by a predetermined calculation process, and a timing at which the occurrence probability is calculated.   A screen display device for displaying a screen having a waveform indicating changes in symptoms of a patient in a psychiatric disease region and a function approximating the waveform. The screen display device according to claim 12,
The screen display apparatus which displays the screen which has the said waveform and the said function before and after prescription of the medicine to the said patient, respectively. A screen generation process for generating a screen having a waveform indicating a change in a patient's symptoms in a predetermined medical treatment area and a function approximating the waveform;
A program executed by an information processing apparatus. The program according to claim 14, wherein
The predetermined medical treatment area is a mental illness area,
A text vectorization process for vectorizing the text information by calculating a frequency of words included in the text information by a predetermined natural language process on the text information described in the medical record in the mental disease region;
An association process for associating a description timing indicating the timing at which the medical record is described with the word and the frequency;
Based on the symptom model including the predetermined word and the frequency of the predetermined word and indicating the characteristic of the symptom of the mental illness region, and the word and the frequency correlated in the association process, An occurrence probability calculation process for calculating an occurrence probability that is unique for each of the described timings;
A waveform acquisition process for acquiring the waveform based on the occurrence probability for each of the described timings calculated by the occurrence probability calculation process;
An approximation process for approximating the acquired waveform with the function;
For causing an information processing apparatus to execute the program.

Images