KR102483538B1 - System and method for predicting adaptability of taking medicine using machine learning - Google Patents

Abstract

The present invention provides a voice counseling service to patients or guardians, and at the same time extracts effective information, which is meaningful information necessary for extracting the patient's medication compliance risk, from voice counseling, and then extracts it using a medication compliance risk extraction algorithm, which is a learned machine learning algorithm. By analyzing valid information and collected my data (financial transactions, payment of communication expenses, etc.), the patient's medication compliance risk is output, and based on the output medication compliance risk information, a medication compliance report is generated and provided to medical staff to help medical staff By performing medical consultation by referring to the medication compliance report, not only can the patient's medication compliance be dramatically increased, but also it is possible to reduce medical cost consumption due to low medication compliance in the past, and counseling concentration, efficiency and It relates to a system and method for predicting medication compliance risk using machine learning, which can maximize the effect and reduce unnecessary budget consumption of insurance companies.

Description

System and method for predicting adaptability of taking medicine using machine learning}

The present invention relates to a system and method for predicting medication compliance risk using machine learning, and more specifically, provides a voice counseling service to a patient or guardian and at the same time provides effective information, which is meaningful information necessary for extracting a patient's medication compliance risk from voice counseling. After extracting, by using the learned machine learning algorithm, the extracted effective information and the collected my data (financial transactions, telecommunications payment, etc.) are analyzed to output the patient's medication compliance risk, and based on the output medication compliance risk information By generating a medication compliance report and providing it to the medical staff, as the medical staff performs medical consultation by referring to the medication compliance report, the patient's medication compliance can be dramatically increased, as well as reducing the consumption of medical costs due to conventional low medication compliance. The present invention relates to a medication compliance risk prediction system and method using machine learning that can maximize consultation concentration, efficiency and effectiveness compared to the same consultation time, and significantly reduce unnecessary budget consumption of insurance companies.

In recent years, as medical standards, medical technology and equipment have developed, and interest in health has soared, medical facilities such as hospitals and pharmacies and the number of users using medical facilities have increased exponentially. Various studies are being conducted to increase reliability.

Usually, medical staff prescribes medicines suitable for the patient's medical history through medical examination of the visiting patients, and even if these prescription drugs are patients with the same disease, the type, dosage, and dosage according to various conditions such as the patient's history, symptoms, and side effects Prescriptions are made with different cycles and concentrations, and the patient is taking the drug according to the doctor's prescription.

At this time, since the treatment effect differs significantly depending on compliance, which indicates how much the patient has taken the prescribed medicine according to the instructions of the medical staff, the medical staff provides detailed instructions to the patient when prescribing medicine, dosage, dosage cycle, instructions and advice, etc. are guiding

However, in the past, due to various causes such as lack of common sense of health, lack of drug knowledge, cognitive decline, complexity of multi-drug prescription, decreased awareness of disease, memory loss, delay in medication due to busy daily life, and lifestyle, patients take medication A problem of very poor compliance occurred.

For example, according to the journal of the Korean Society of Information Technology in July 2020, 'Medicine compliance in major developed countries accounts for only 50% of the total drug-taking population, 77% in diabetic patients, and 45% in hypertension patients. , the medication compliance of patients with hyperlipidemia is less than 35%.'

In other words, although medication compliance is a prerequisite for patient treatment success, about half of all patients still have low medication compliance, and accordingly, the hospital admission rate and mortality rate increase, resulting in an increase in overall medical expenses.

In order to improve the patient's medication compliance, first of all, 1) the medical staff must accurately recognize the patient's current medication compliance status and give intensive instructions and guidance on the insufficient part during medical consultation, and 2) the medical staff Patients should be provided with detailed medication information such as type of prescription, dose, dosage cycle, instructions and advice.

However, in contrast to the limited medical facilities, as the number of users using the medical facilities increases, the medical staff's treatment hours per patient are limited, and the medical staff accurately recognizes the patient's medication compliance status during the short consultation time or In reality, it is not easy to provide detailed information on medications.

In other words, it is very urgent to study to improve medication compliance, which is directly related to patient treatment success and medical costs, but research on this is still very insufficient.

In general, insurance companies provide compensation for incurred medical expenses (medical treatment, prescription, etc.) only for patients enrolled in insurance products. Conventionally, insurance companies do not have separate technologies and methods for recognizing patients' medication compliance. In addition, since insurance premiums are set and insurance money is paid equally to all patients regardless of the patient's medication compliance, unnecessary budget consumption by insurers increases and the patient's willingness to improve medication compliance is reduced. Occurs. For example, assuming that insurance companies set insurance premiums or pay insurance premiums differently depending on patient compliance, patients voluntarily work to improve their medication compliance, and this improvement in medication compliance consumes the insurance company's budget. It not only reduces the cost of treatment, but also creates the effect of increasing the patient treatment rate.

1 is a block diagram showing the oral medication compliance monitoring system disclosed in Korean Patent Registration No. 10-1510512 (Title of Invention: Oral Medication Compliance Monitoring System and Method Using Personal Identification RFID Tag).

The oral medication compliance monitoring system (hereinafter referred to as the prior art) 100 of FIG. 1 includes the name of the oral medication, the medication method, the medication period, the number of medications taken per day and the medication time, precautions when taking medication, side effects of the medication, and regular medication prescription. A memory module 110 that stores information on whether or not the oral drug is present in advance, an RFID tag 120 attached to the surface of the oral drug and storing unique information of the oral drug and DNA information for user identification, and an RFID tag before taking the oral drug ( 120) to block RF radiation and to be removed by gastric acid when the user takes oral medication and the oral medication reaches the stomach, and the RFID protective film is removed to provide unique information and user When an RFID signal including DNA information for identification is radiated, the RF module 140 that receives the RFID signal including the unique information of the oral drug and the DNA information for user identification and the RFID signal received by the RF module 140 and a medication compliance confirmation module 170 that continuously monitors the user's oral medication intake by processing and recognizing unique information of the oral medication and DNA information for user identification.

At this time, DNA information for user identification applied to the medication compliance confirmation module 170 is a STR marker (short tandem repeat marker) that can identify an individual, and the RFID protective film is removed by an acid with a pH of 1.5 to 3.5. It is configured to have heat resistance that does not melt in the heat of the temperature range and water resistance that does not dissolve in water.

In addition, the medication compliance confirmation module 170 is configured to automatically transmit an automatic text message or automatic call information to an emergency center or police station as a result of oral medication monitoring and, if the oral medication is a ready-made sleeping pill that has a risk of death in case of an overdose at once. do.

In addition, the medication compliance confirmation module 170 analyzes whether the user takes the oral medication according to the prescribed medication method at the medication time within the medication period previously stored in the memory module 110, and as a result of the analysis, does not take the oral medication at the prescribed medication time or When there is an abnormality in taking oral medication, the user is notified through the display module 150, and an alarm message is transmitted to the personal doctor's portable terminal 180 through the communication module 160 or analyzed by the server 190 of the hospital. It is configured to back up the results.

The prior art 100 configured as described above has an advantage of being able to objectively monitor whether or not the user is taking the oral medication according to an accurate medication intake method by attaching an RFID tag to the oral medication.

However, since the prior art 100 simply uses the RFID tag 120 to monitor whether or not to take the drug, it has a structural limitation in not increasing the patient's voluntary participation in taking the drug.

In addition, since the prior art 100 requires a configuration for attaching an RFID tag to an oral drug and a configuration for removing the RFID protective film by an acid of pH 1.5 SOWL 3.5, manufacturing of prescription drugs is not only complicated and cumbersome, but rather expensive. This growing problem arises.

In addition, the prior art 100 has a disadvantage in that it is difficult to apply to injections widely used for insulin preparations or autoimmune diseases that require self-injection by patients themselves. Biological agents prepared in the form of such injections show a continuous increase in terms of new drug releases and usage.

In general, even if the patient takes the drug exactly as suggested by the medical staff, different effects and side effects may occur depending on various conditions such as the patient's (user's) physical characteristics, physical condition, and life pattern, and such a patient's condition The best way to accurately acquire is to obtain various information about the patient through consultation with the patient.

However, the prior art 100 is configured to simply recognize whether or not an oral drug is taken, and it is impossible to obtain information on the effects and side effects of the patient's drug intake, and accordingly, the medical staff monitors the patient's condition in detail during the limited consultation time. However, it is difficult to identify accurately and has the disadvantage of lowering the effectiveness and efficiency of treatment.

In particular, in the case of patients with cognitive decline, such as old age, brain surgery, and mental illness, conversation and communication with the patient are more difficult due to reasons such as not giving accurate answers to the doctor's questions or having free conversation with the doctor's questions. Because of the difficulty, the efficiency of medical consultation is further reduced.

The present invention is to solve this problem, and the problem of the present invention is to provide a voice counseling service to the patient or guardian, and at the same time extract effective information, which is meaningful information necessary for extracting the patient's medication compliance risk, from the voice counseling. Using the medication compliance risk extraction algorithm, which is a machine learning algorithm, the extracted effective information and the collected my data (financial transactions, communication expenses, etc.) are analyzed to output the patient's medication compliance risk, based on the output medication compliance risk information. By generating a medication compliance report and providing it to the medical staff, as the medical staff performs medical consultation by referring to the medication compliance report, the patient's medication compliance can be dramatically increased, as well as reducing the consumption of medical costs due to conventional low medication compliance. The purpose of this study is to provide a medication compliance risk prediction system and method using machine learning that can maximize consultation concentration, efficiency and effectiveness compared to the same consultation time.

In addition, another problem of the present invention is that the patient's recurrence or deterioration is reduced as the counseling service management server provides voice counseling service with the patient, which can significantly reduce the patient's treatment cost, as well as the patient's time and space constraints. It is to provide a system and method for predicting medication adherence risk using machine learning that can increase medical convenience by enabling medical consultation without

In addition, another problem of the present invention is to improve the accuracy and reliability of the patient's medication compliance risk extraction by using the collected my data as input data of the medication compliance risk extraction algorithm after the counseling service management server collects the patient's My Data. It is to provide a medication adherence risk prediction system and method using machine learning that can further increase

In addition, another problem of the present invention is to individually proceed with supervised learning and unsupervised learning at the same time when the medication compliance risk extraction algorithm learns, so as to compensate for the disadvantages of the model according to each method. It is intended to provide a medication compliance risk prediction system and method using machine learning that can more precisely extract the medication compliance risk by being configured.

In addition, another problem of the present invention is that the consultation level setting unit calculates the consumption time (t2) consisting of meaningless information among the patient's voice data, and then calculates the consultation level (L) of the patient according to the calculated consumption time (t2). It is to provide a medication compliance risk prediction system and method that can further increase the efficiency and effectiveness of medical consultation by detecting and AI chatbot proceeding with different difficulty or speed of conversation for each consultation grade (L).

In addition, another problem of the present invention is to refer to the medication compliance report transmitted by the insurance company server from the counseling service management server, to differentiate insurance premiums according to the risk of medication compliance when purchasing insurance, or to compensate for insurance money, according to the risk of medication compliance. It is to provide a medication compliance risk prediction system and method that can dramatically reduce unnecessary budget consumption of insurance companies by paying insurance money differently and increase patient's willingness and participation rate for medication compliance improvement to increase disease treatment rate.

The solution of the present invention for solving the above problems is a medication compliance risk prediction system including a patient terminal and a consultation service management server: the consultation service management server includes check items necessary for extracting the patient's medication compliance risk, and each check Query words for inducing answers to items, keywords that can be judged as words related to each check item are preset and stored, and valid information for extracting valid information, which is meaningful information necessary for extracting the risk of medication compliance from the patient's corpus, is stored. a database unit storing an extraction algorithm and a medication compliance risk extraction algorithm that analyzes the extracted effective information and outputs a medication compliance risk of the corresponding patient; It is executed every predetermined period (T), generates learning data capable of learning the correlation between the effective information and the risk of medication compliance, and utilizes the generated learning data to obtain parameter values for the correlation between the effective information and the risk of medication compliance. an artificial intelligence learning unit that performs learning of the medication compliance risk extraction algorithm by deriving an extraction model which is a set; a voice counseling service unit that connects to the patient-side terminal and executes voice counseling at the request of the patient-side terminal; After recognizing the patient's voice data input from the voice counseling service unit, generating a corpus, valid information is extracted from the generated patient corpus using the valid information extraction algorithm, and the extracted valid information is an artificial intelligence-based medication adherence risk extraction unit that outputs a medication compliance risk of the patient based on input data of the medication compliance risk extraction algorithm learned by the artificial intelligence learning unit; a medication compliance report generation unit generating a medication compliance report based on the medication compliance risk information extracted by the artificial intelligence-based medication compliance risk extraction unit; It is executed when the voice counseling service ends, and when the patient's voice data executed through the voice counseling service unit is input, the inputted patient's voice data is analyzed according to a preset rating procedure to communicate with the patient. After setting a consultation level (L) according to the ability level, a consultation level setting unit for storing in the database unit; A control unit for transmitting the medication compliance report generated by the medication compliance report generating unit to a medical staff terminal of the corresponding patient, and the voice counseling service unit may store the counseling level (L) of the patient requesting counseling in the database unit. At this time, the difficulty or speed of conversation is different for each consultation level (L) of the patient, and the consultation level setting unit uses the patient's voice data executed through the voice consultation service unit and the artificial intelligence-based medication compliance risk extraction unit. a data input module for receiving valid information extracted by; Voice data input through the data input module is sorted into a bit string according to the timeline, and the input valid information is referenced and utilized to output the valid information output section and valid information for the voice data arranged in the bit string. a voice data classification module for classifying into sections that do not correspond; a total output time (t) calculation module that calculates a total output time (t), which is a total output time of the audio data input by the data input module; a consumption time (t2) calculating module that calculates a consumption time (t2), which is an output time obtained by adding up sections in which valid information is not output by the voice data classification module; Consumption rate (R) that calculates the consumption rate (R), which is a percentage of the consumption time (t2) calculated by the consumption time (t2) calculation module and the total output time (t) calculated by the total output time (t) calculation module calculation module; Using the reference table in which the counseling grade (L) is matched for each section of the consumption rate (R), the counseling grade (L) corresponding to the consumption rate (R) calculated by the consumption rate (R) calculation module is detected, and the detected After determining the counseling grade (L) as the counseling grade (L) of the patient, it includes a counseling grade determination module for storing in the database unit.

In addition, in the present invention, the medication compliance risk prediction system is for the patient requested from the counseling service management server, past health examination results, account information, loan payment information, automobile insurance contract and payment information, insurance loan product information, card use information , Communication fee payment information, guarantee insurance information, national tax payment information, local tax payment information, and my data collection server for collecting at least one or more of the 4 major insurance payment information, wherein the consultation service management server further includes the above It is executed when voice counseling is started by the voice counseling service unit, and after requesting My Data for the corresponding patient from the My Data collection server, the My Data collection unit receives My Data of the corresponding patient from the My Data collection server. And, the medication compliance risk extraction algorithm uses the effective information extracted by the artificial intelligence-based medication compliance risk extraction unit and My Data collected by the My Data collection server as input data to output the medication compliance risk desirable.

In addition, in the present invention, the artificial intelligence learning unit simultaneously and individually proceeds with supervised learning and unsupervised learning when learning the medication compliance risk extraction algorithm, mutually complementing the disadvantages of the models according to each method. It is desirable to do

In addition, in the present invention, the artificial intelligence-based medication compliance risk extraction unit includes a voice data input module that receives the patient's voice data from the voice counseling service unit; My data input module receiving my data from the my data collection unit; a voice recognition and corpus generating module that recognizes voice by analyzing the voice data input by the voice data input module and generates a corpus of the recognized patient's voice data; a valid information extraction module for extracting valid information from the patient corpus by analyzing the patient corpus generated by the voice recognition and corpus generating module using the valid information extraction algorithm; When valid information and my data are received from the valid information extraction module and the my data input module, the input valid information and my data are used as input data of the learned medication compliance risk extraction algorithm to determine the medication compliance risk for the patient It is preferable to include an AI-based medication compliance risk output module.

In addition, in the present invention, the medication compliance risk prediction system further includes an insurance company server, the insurance company server provides pre-designed insurance products, and requests a report including identification information of the patient when a patient subscribes to an insurance product or pays compensation. After the data is transmitted to the counseling service management server, the insurance premium of the subscribed insurance product or compensation is differentially paid according to the medication compliance risk of the medication compliance report transmitted from the counseling service management server, and the counseling service management When receiving the report request data from the insurance company server, the server preferably transmits a medication compliance report of the corresponding patient to the insurance company server.

In addition, in the present invention, the medication compliance risk output by the medication compliance risk extraction algorithm includes the medication compliance of the patient and medication compliance grades for each preset category, and the medication compliance report shows the medication compliance of the patient as a percentage. At the same time as displaying, it is desirable to display medication compliance grades of each category.

In addition, in the present invention, a consultation application is installed in the patient-side terminal, and when a consultation request is received from a user (patient), the consultation application transmits consultation request data to the consultation service management server and the consultation with the user (patient) at the same time. It is preferable to display a counseling interface, which is a graphical user interface (GUI) providing an interface for the voice counseling service of the service management server, on the monitor of the patient-side terminal.

In addition, when the consultation application in the present invention receives basic information including personal information, type of disease, information of medical staff who performed previous medical treatment, and information of medical staff who wish to receive medical treatment from the user (patient) when first executed, the input basic information is transmitted to the consultation service management server, and when the consultation service management server receives basic information from the consultation application, the received basic information is stored in the database unit, and at the same time, medical staff or medical staff who performed previous medical treatment included in the basic information It is preferable to further include a medical staff matching unit that matches a medical staff who wishes to treat the patient with a medical staff who will treat the patient.

delete

delete

In addition, in the present invention, the database unit further stores a second reference table in which the ventilation period output period (T1) for each consultation level (L) is matched, and the ventilation period output period (T1) of the second reference table is the consultation level ( Assuming that the higher L) is the upper level (or lower level), it is set in proportion to (or inversely proportional to) the consultation level L, and the consultation service management server determines the consultation level by the consultation level setting unit. Executed, using the second reference table, after detecting the output period (T1) of the ventilation message that matches the consultation level (L) of the patient determined by the consultation level setting unit, and then storing the ventilation statement in the database unit an output period (T1) detector; The voice counseling service unit further includes an output determination unit that is executed when the voice consultation starts and terminates when the voice consultation ends, and the determination unit searches the database unit when first executed, A data search and extraction module for searching for a ventilation comment output period (T1) for the patient; a monitoring module that is executed when the output cycle (T1) of the patient is extracted by the data search and extraction module, and monitors the voice counseling service unit and the effective information extraction module; An elapsed time (Δt) calculation module for calculating an elapsed time (Δt) from the time when the voice counseling service unit asks the query to the present in which a state in which valid information is not extracted continues, when monitored by the monitoring module; a comparison module for comparing the elapsed time (Δt) calculated by the elapsed time (Δt) calculation module with the excitation output period (T1) extracted by the data search and extraction module; In the comparison module, if the elapsed time (Δt) is more than the ventilation comment output period (T1) (Δt≥T1), the ventilation comment is reached (Δt < T1), after determining that the AI chatbot should output the ventilation comment , an arousal message output determining module for outputting an output control signal to the voice counseling service unit, and when the output control signal is input from the arousal message output determination unit, the voice counseling service unit cuts off the patient's conversation and repeats the previous query. It is desirable to output an arousal message meaning a repeated comment to ask.

According to the present invention having the above problems and solutions, a voice counseling service is provided to the patient or guardian, and at the same time, effective information, which is meaningful information necessary for extracting the patient's medication compliance risk, is extracted from the voice counseling, and then the learned machine learning algorithm, medication compliance The effective information extracted using the risk extraction algorithm and the collected my data (financial transaction, telecommunications payment, etc.) are analyzed to output the patient's medication compliance risk, and a medication compliance report is generated based on the output medication compliance risk information. By providing this to the medical staff, as the medical staff conducts medical consultation by referring to the medication compliance report, the patient's medication compliance can be dramatically increased, and the medical cost consumption due to the conventional low medication compliance can be reduced, and the same treatment You will be able to maximize counseling concentration, efficiency and effectiveness compared to counseling time.

In addition, according to the present invention, as the counseling service management server provides a voice counseling service with the patient, the recurrence or deterioration of the patient is reduced, thereby significantly reducing the patient's medical treatment cost, and the patient can receive medical counseling without time and space limitations. This can increase the convenience of treatment.

In addition, according to the present invention, after the counseling service management server collects the patient's My Data, the collected My Data is used as input data of the medication compliance risk extraction algorithm, thereby further increasing the accuracy and reliability of the patient's medication compliance risk extraction. there will be

In addition, according to the present invention, when the medication compliance risk extraction algorithm is trained, supervised learning and unsupervised learning are conducted individually at the same time, and the medication compliance is configured to compensate for the disadvantages of the model according to each method. Risk can be extracted more precisely.

In addition, according to the present invention, after calculating the consumption time (t2) consisting of meaningless information among the patient's voice data, the consultation level setting unit detects the patient's consultation level (L) according to the calculated consumption time (t2), AI As the chatbot proceeds with different levels of difficulty or speed of conversation for each consultation level (L), the efficiency and effectiveness of medical consultation can be further increased.

In addition, according to the present invention, the insurance company server refers to the medication compliance report transmitted from the counseling service management server to differentiate insurance premiums according to the risk of medication compliance when purchasing insurance, or to pay insurance money differently according to the risk of medication compliance when insurance money is compensated. By doing so, it is possible to drastically reduce unnecessary budget consumption by insurance companies and increase the rate of disease treatment by increasing patients' willingness and participation rate in improving medication adherence.

1 is a block diagram showing the oral medication compliance monitoring system disclosed in Korean Patent Registration No. 10-1510512 (Title of Invention: Oral Medication Compliance Monitoring System and Method Using Personal Identification RFID Tag).
2 is a block diagram showing a system for predicting medication compliance risk according to an embodiment of the present invention.
FIG. 3 is a relationship diagram of FIG. 2 .
4 is an exemplary diagram illustrating a medication compliance report generated by the counseling service management server of FIG. 2 .
5 is another exemplary view of FIG. 4 .
6 is a block diagram illustrating a consultation service management server of FIG. 2 .
FIG. 7 is a conceptual diagram for explaining FIG. 6 .
8 is a block diagram illustrating an AI chatbot-based voice consultation service unit of FIG. 6 .
9 is a block diagram illustrating an artificial intelligence-based medication compliance risk extraction unit of FIG. 6 .
10 is a flowchart for explaining an operation process of a medication compliance risk prediction system according to an embodiment of the present invention.
FIG. 11 is a block diagram illustrating a second embodiment of a counseling service management server of FIG. 6 .
FIG. 12 is a block diagram illustrating a determination unit for outputting a ventilation message in FIG. 11 .
FIG. 13 is a block diagram illustrating a consultation level setting unit of FIG. 11 .
FIG. 14 is an exemplary diagram for explaining the total output time (t), valid time (t1), and consumption time (t2) of FIG. 13 .
Fig. 15 is a flowchart showing the second embodiment of Fig. 10;

Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.

2 is a configuration diagram showing a system for predicting medication compliance risk according to an embodiment of the present invention, and FIG. 3 is a relationship diagram of FIG. 2 .

The medication compliance risk prediction system 1, which is an embodiment of the present invention, provides a voice counseling service to a patient or guardian, extracts effective information, which is meaningful information necessary for extracting the patient's medication compliance risk, from the voice counseling, and then learns machine learning. By analyzing the effective information extracted using an algorithm and the collected my data (financial transaction, communication fee payment, etc.), the patient's medication compliance risk information is output, and a medication compliance report is generated based on the output medication compliance risk information. By providing this to the medical staff, as the medical staff conducts medical consultation by referring to the medication compliance report, the patient's medication compliance can be dramatically increased, and medical costs due to conventionally low medication compliance can be reduced, and the same medical consultation It is to maximize counseling concentration, efficiency and effectiveness in relation to time.

In addition, as shown in FIGS. 2 and 3, the medication compliance risk prediction system 1 of the present invention provides a counseling service to the connected user and measures the medication compliance risk from voice data through voice counseling with the connected user. After extracting meaningful information (hereinafter referred to as effective information) related to extraction, extracting the risk of medication compliance by analyzing the extracted effective information using a machine learning algorithm, and generating a medication compliance report based on this, a counseling service management server (3) ), patient-side terminals 5-1, ..., (5-N), which are terminals owned by the patient or patient guardian, medical staff terminal 9, which is a terminal owned by a doctor, and patient-side terminals 5-N. 1), ..., (5-N) installed in each of the consultation application (7) to support interworking with the consultation service management server (3), and a specific patient according to the request of the consultation service management server (3) A My Data collection server 11 that collects and provides My Data (financial transaction information, telecommunication fee payment history information, etc.) and an insurance company server 13 that provides insurance products and supports the insurance premium setting and insurance payment of the subscribed patient. And, a communication network that provides a data movement path between the medical staff terminal 9, my data collection server 11, insurance company server 13, and patient terminals 5-1, ..., (5-N) ( 10).

At this time, in the present invention, for convenience of description, the patient-side terminal 5 and the consultation service management server 3 have been described as an example in which voice counseling is performed through a separate consultation-only application 7, but the patient-side terminal 5 And the voice counseling method of the counseling service management server 3 is not limited to this, and it is natural that it can be configured to be operated based on voice calls through a normal telephone network.

In addition, in the present invention, for convenience of explanation, it has been described as an example that an AI chatbot performs a consultation service with a patient using artificial intelligence (AI), but the consultation service of the consultation service management server 3 This can be done through a phone call with a professional counselor.

The communication network 10 includes a counseling service management server 3, medical staff terminal 9, my data collection server 11, insurance company server 13, and patient terminals 5-1, ..., (5-N). It supports data communication between devices, and in detail, LAN (local area network) network, VAN (Value Added Network) network, wide area network (WAN), mobile communication network, Wi-Fi, 3G/4G, LTE etc. can be implemented.

The My Data collection server 11 collects information (hereinafter referred to as My Data) such as the patient's loan payment details, insurance premium payment details, and communication cost payment details. At this time, the My Data collection server 11 collects My Data only from patients who agree to collect My Data.

For example, My Data includes 'past health checkup results', 'account information', 'loan payment information', 'auto insurance contract and payment information', 'insurance loan product information', 'card usage information (daily card usage information). Included', 'card loan information', 'other financial investment information', 'communication fee payment information', 'guarantee insurance information', 'public information (national tax, local tax, 4 major insurances)', etc. Since the collection technology and method are commonly and widely used in financial transaction systems, a detailed description thereof will be omitted.

In addition, the My Data collection server 11 extracts My Data of a specific patient according to the request of the consultation service management server 3 and transmits the extracted My Data to the consultation service management server 3 .

Such My Data can be used as an indicator of a patient's lifestyle or economic situation, and in general, a patient's lifestyle or economic situation has a close relationship with the patient's adherence to medication. For example, in the case of a patient who has a lifestyle of overdue monthly expenses, the cycle of taking the drug is often delayed, and when the economic situation is very poor, the continuity of the medication is often poor.

In the present invention, considering the characteristics of My Data, the counseling service management server 3 separately collects My Data of the patient, and then uses the collected My Data to extract the patient's medication compliance risk, thereby increasing the accuracy and reliability of the analysis. was significantly increased.

The medical staff terminal 9 is a terminal carried by a medical staff or disposed in the medical staff's office space, and in detail, a desktop-PC, a smart phone, a laptop computer, a tablet-PC ( Tablet-PC) and the like. In other unavoidable cases, it is natural that a document in a printed form may also be configured.

In addition, when the medical staff terminal 9 receives a medication compliance report of a specific patient from the counseling service management server 3, it displays the received medication compliance report on the monitor at the request of the user (medical staff).

In general, during the limited medical consultation time, the medical staff needs to find out whether the patient has taken the medicine according to the instructions and guidance of the medical staff, so not only does it carry out a conversation with the patient, but also if the patient has high medication non-compliance, the cause Efforts (conversation, guidance, instructions, etc.) are being made to increase patient adherence to medication by identifying

In particular, since the patient's medication compliance has a characteristic that increases as the consultation time with the medical staff increases, the medical staff needs sufficient medical consultation time. There was a problem in that it was impossible to accurately grasp the prescription drug, provide sufficient explanation about the prescription drug, or guide the patient to improve medication compliance.

That is, the present invention is to solve this problem, and the medical staff refers to the contents of the medication compliance report transmitted from the consultation service management server 3 through the medical staff terminal 9 to obtain basic information related to the patient's medication compliance. In-depth analysis of the patient is possible as intensive counseling can be conducted on the part lacking in the patient or the part requiring in-depth analysis during the limited medical consultation time. Conversation) can be significantly reduced to significantly increase the patient's medication compliance, counseling efficiency, and treatment efficiency, thereby effectively reducing the consumption of medical expenses.

The patient-side terminals 5-1, ..., and 5-N are digital devices possessed by patients or patient guardians and supporting connection with the communication network 10, and in detail, smart phones. , Desktop-PC (Desktop-PC), notebook (Notebook), tablet-PC (Tablet-PC) can be configured.

In addition, the patient-side terminals 5-1, ..., 5-N include input means 51 such as a keyboard or touch panel for receiving text from the user, and audio output means such as a speaker for outputting sound ( 52), voice collection means 53 such as a microphone for collecting voice signals, display means 54 displaying text, numbers, photos, video contents, etc., and communication supporting connection with the communication network 10 Including means 55 to support voice counseling with an AI chat-bot of the consultation service management server 3 of FIG. 5 to be described later.

In addition, a counseling application 7 described later is installed in the patient-side terminals 5-1, ..., and 5-N.

The counseling application (7) is installed on the patient-side terminal (5) and interworks with the counseling service management server (3) to provide a user (patient) with a voice consultation service with an AI chatbot. Software, application, or It is an application program.

In addition, when the counseling application 7 is first executed, basic information is received from the patient (User). In this case, the basic information may include at least one of the patient's personal information, the type of disease, information on medical facilities and medical staff who have performed previous medical treatment, and information on medical facilities and medical staff who wish to receive medical treatment.

In addition, when receiving basic information from a patient, the consultation application 7 transmits the input basic information to the consultation service management server 3 . At this time, the consultation service management server 3 matches the medical staff terminal 9 to treat the patient based on the basic information transmitted from the consultation application 7 .

In addition, when the consultation application (7) receives a voice consultation request from the patient, it transmits consultation request data to the consultation service management server (3) and at the same time provides a graphic for performing voice consultation with the AI chatbot of the consultation service management server (3) An interface for counseling, which is a GUI (Graphic User Interface), is displayed on the display unit 54 of the patient-side terminal 5. At this time, the consultation interface provides an interface for voice consultation service between the user (patient or patient guardian) and the AI chatbot of the consultation service management server (3).

The insurance company server 13 is a server of a general insurance company that provides various insurance products, and provides a service of setting insurance premiums for patients subscribed to insurance products and paying insurance money upon compensation.

In this case, the insurance premium refers to the cost that the subscriber pays to the insurance company every month, and the insurance money refers to the compensation paid by the insurer to the subscriber.

In addition, the insurance company server 13 transmits report request data including identification information (personal information, etc.) of the patient to the consultation service management server 3 when a patient is registered or compensated. At this time, upon receiving report request data from the insurance company server 13, the counseling service management server 3 transmits a medication compliance report of the corresponding patient to the insurance company server 13.

In addition, when the insurance company server 13 receives the medication compliance report from the counseling service management server 3, 1) at the time of patient subscription, the insurance premium is differentiated according to the medication compliance risk of the medication compliance report, 2) when the patient is compensated, Depending on the risk of medication compliance in the medication compliance report, insurance money is paid differently.

That is, the present invention refers to the medication adherence report transmitted from the counseling service management server 3 when the insurance company server 13 registers or compensates for a patient, and differentially applies insurance premiums or insurance money payments according to the risk of medication adherence, thereby avoiding the insurer's unnecessary In addition to drastically reducing budget consumption, it is possible to increase the patient's willingness and participation rate in improving medication compliance, thereby increasing the disease treatment rate.

The counseling service management server 3 is a server for providing counseling services to patients or their guardians and at the same time providing medical staff with a medication adherence report including a patient's medication adherence risk.

In addition, when the consultation service management server 3 receives basic information from the consultation application 7, it stores the received basic information and at the same time the contents included in the basic information (type of disease, medical facilities and medical staff who performed previous treatment) information, medical facility and medical staff information) to match the doctor who treated the patient.

At this time, the matching method may be operated in such a way that the medical facility and medical staff desired by the patient are selected, or the medical facility and medical staff previously treated by the patient are selected. 31) is stored.

In addition, when the consultation service management server 3 receives a consultation request from the consultation application 7, it executes an artificial intelligence-based AI chatbot and performs voice consultation with the corresponding patient through the AI chatbot.

At this time, in the present invention, for convenience of explanation, it has been described that voice counseling with a patient is performed by an AI chatbot as an example, but voice counseling of the counseling service management server 3 can be performed through a call with a normal professional counselor .

Table 1 below is a table showing basic check items and query words that can be grasped through the AI chatbot or questionnaire of the consultation service management server of FIG. 2.

[Table 1]

The basic check items in Table 1 refer to check items that are basically necessary when medical staff consults patients, and the query words of each basic check item refer to questions to derive information about the corresponding basic check item from the patient.

In addition, these basic check items may be obtained through voice counseling of the AI chatbot of the consultation service management server 3 of FIG. 2 or obtained through a medical examination table (preliminary examination table).

That is, the AI chatbot of the counseling service management server 3 conducts a daily conversation with the patient, but as shown in Table 1, queries for inducing information on basic check items from the patient are naturally provided to the patient according to the situation. Ask a question.

Table 2 below is a table showing essential check items and query words that can be grasped through voice counseling of the AI chatbot of the consultation service management server of FIG. 2 .

[Table 2]

The essential check items in Table 2 refer to check items required to extract the patient's medication compliance risk, and the query words of each basic check item refer to questions for inducing information on the corresponding essential check items from the patient.

In addition, these essential check items are obtained through voice consultation of the AI chatbot of the consultation service management server (3).

That is, the AI chatbot of the counseling service management server 3 conducts a daily conversation with the patient, but as shown in Table 2, queries for inducing information on essential check items from the patient are naturally provided to the patient according to the situation. Ask a question.

At this time, as the essential check items are configured differently according to the patient's symptoms and medical history, the AI chatbot applied to the consultation service management server (3) of the present invention includes essential check items according to the patient's symptoms and medical history and accordingly Query terms are previously set (set) and stored.

At this time, essential check items may be set by receiving information in advance from experts and medical staff.

In addition, keywords that can be determined as words related to each essential check item from the patient's conversation are preset and stored in the counseling service management server 3 .

At this time, since the essential check items may vary according to the type of disease, medical history and symptoms of the patient, the keywords may also consist of different configurations depending on the type of disease, medical history and symptoms of the patient, and, like the essential check items, experts and medical staff It may be set by receiving information in advance.

In addition, when the voice counseling of the AI chatbot starts, the consultation service management server (3) determines whether the patient agrees to collect My Data or not, and if the patient agrees to collect My Data, If , My Data is requested from the My Data collection server 11, and My Data of the corresponding patient is received from the My Data collection server 11.

In addition, when the voice counseling of the AI chatbot starts, the counseling service management server (3) extracts valid information (keywords and sentences) related to basic check items and essential check items from the patient's corpus during voice counseling. After analyzing the extracted effective information and the collected my data using the extracted medication compliance risk extraction algorithm, the medication compliance risk of the patient is extracted, and then a medication compliance report is generated using the extracted medication compliance risk.

At this time, the medication compliance risk extraction algorithm is a machine learning algorithm that extracts the patient's medication compliance risk using valid information and my data as input data, and the learning and analysis methods will be described in detail in FIGS. 6 and 9 described later. .

In addition, when the medication compliance report is generated, the counseling service management server 3 transmits the generated medication compliance report to the corresponding medical staff terminal 9 . At this time, various known methods such as encrypted e-mail, text message, dedicated program, and mail delivery may be applied to the method of transmitting the medication compliance report of the counseling service management server 3 .

In addition, upon receiving report request data from the insurance company server 13, the counseling service management server 3 transmits a medication compliance report for the corresponding patient to the insurance company server 13.

4 is an exemplary diagram illustrating a medication compliance report generated by the counseling service management server of FIG. 2 , and FIG. 5 is another exemplary diagram of FIG. 4 .

As shown in FIGS. 4 and 5, the medication compliance report 900 displays the patient's medication compliance as a percentage, and includes 'patient's condition', 'guardian's disease management knowledge', 'guardian's patient control', ' Medication compliance grades (very good, average, bad, very bad) are displayed for each preset category, such as 'reliability in medical system' and 'lifestyle based on my data'.

At this time, in the present invention, the format and configuration of the medication compliance report 900 has been described as an example with FIGS. 4 and 5, but the format and configuration of the medication compliance report 900 is not limited thereto, and may be made in various formats and configurations. .

That is, as the doctor reads and refers to the medication compliance report 900 transmitted from the counseling service management server 3 before actual medical consultation with the corresponding patient and performs medical consultation, the medication compliance report 900 is insufficient. It is possible to consult intensively on items or items requiring in-depth analysis, and accordingly, the effect of diagnosis and prescription can be increased by maximizing the patient's medication compliance compared to the limited consultation time.

In addition, the insurance company server 13 refers to the medication compliance report 900 transmitted from the counseling service management server 3 when signing up for a patient or paying insurance money, and differentiates insurance premiums or insurance money according to the risk of medication compliance in the medication compliance report. In addition to drastically reducing unnecessary budget consumption by insurers, it is possible to increase the rate of disease treatment by increasing patients' willingness and participation rate in improving medication adherence.

FIG. 6 is a block diagram illustrating the consultation service management server of FIG. 2, and FIG. 7 is a conceptual diagram for explaining FIG.

As shown in FIGS. 6 and 7, the consultation service management server 3 includes a control unit 30, a database unit 31, a communication interface unit 32, an artificial intelligence learning unit 33, and a medical staff matching unit 34. ), AI chatbot-based voice counseling service unit 35, my data collection unit 36, artificial intelligence-based medication compliance risk extraction unit 37, medication compliance report generation unit 38, and additional service unit 39 .

The control unit 30 is an O.S (Operating System) of the consultation service management server 3, and the control target 31, (32), (33), (34), (35), (36), (37), (38), (39) are managed and controlled.

In addition, the control unit 30 executes the artificial intelligence learning unit 33 at every predetermined period (T).

In addition, when receiving basic information from the counseling application 7 connected through the communication interface unit 32, the control unit 30 stores the received basic information in the database unit 31 and inputs it to the medical staff matching unit 34 at the same time. And, the matching information generated by the medical staff matching unit 34 is stored in the database unit 31.

In addition, when receiving consultation request data from the consultation application 7 through the communication interface unit 32, the control unit 30 executes the AI chatbot-based voice consultation service unit 35.

In addition, when voice counseling is started by the AI chatbot-based voice counseling service unit 35, the control unit 30 sets the my data collection unit 36 and the artificial intelligence-based medication adherence risk extraction unit 37 to the My Data collection unit 36 ) and inputs the collected my data to the artificial intelligence-based medication compliance risk extraction unit 37.

In addition, when the voice counseling is terminated by the AI chatbot-based voice counseling service unit 35, the control unit 30 converts the medication compliance risk output by the artificial intelligence-based medication compliance risk extraction unit 37 to the medication compliance report generator 38 ) is entered.

Also, when the medication compliance report is generated by the medication compliance report generator 38, the controller 30 controls the communication interface 32 to transmit the generated medication compliance report to the medical staff terminal 9.

In addition, when the report request data is received from the insurance company server 13, the control unit 30 extracts the patient's medication compliance report and transmits the extracted medication compliance report to the insurance company server 13 through the communication interface unit 32. Let it be.

The database unit 31 stores basic information of each patient received from each counseling application 7 .

In addition, the database unit 31 stores matching information of patients and medical staff generated by the medical staff matching unit 34 .

In addition, basic check items and essential check items for each disease type, queries of each basic/required check item, and keywords of each basic/required check item are preset and stored in the database unit 31. At this time, keywords are defined as words that can be determined as conversations related to each basic/required check item.

In addition, the database unit 31 stores the medication compliance risk extraction algorithm learned by the artificial intelligence learning unit 33 .

At this time, the medication compliance risk extraction algorithm is learned by the artificial intelligence learning unit 33 at each predetermined cycle (T), and the valid information extracted from the patient's corpus and my data collected by the my data collection unit are used as input data It is an artificial intelligence algorithm that outputs the patient's medication compliance risk.

In addition, the database unit 31 stores an AI chatbot that connects to the counseling application 7 that requested counseling and performs an artificial intelligence-based voice counseling service with the patient.

These AI chatbots utilize deep-learning algorithms to perform voice conversations with patients connected based on artificial intelligence, and at the same time listen to effective information from patients by asking basic/necessary check items and questions according to the type of disease of the patient. Induce a conversation for the purpose, and in detail, by using a deep-learning algorithm for speech recognition, natural language processing (NLP), etc., the input sentence and the output sentence are paired and the most appropriate expression result is extracted. . At this time, NLP (Natural Language Processing) uses five algorithms of word embedding: CBoW (Continuous Bag-of-Words), RN (Relation Network), CNN (Convolutional Neural Network), Self-Attention, It was conducted through a combination of RNN (Recurrent Neural Network), and for sentiment analysis (Opinion Mining), SentiWordNet vocabulary dictionary (positive, negative, and neutral polarity of words) and NLTK (Natural Language Toolkit, natural language) A semi-supervised learning algorithm using a library of processing) was also utilized.

In addition, the database unit 31 stores voice data, which is the contents of the chatbot and the patient's voice consultation performed by the AI chatbot-based voice consultation service unit 35.

In addition, the medication compliance report generated by the medication compliance report generating unit 38 is stored in the database unit 31 .

The communication interface unit 32 transmits and receives data with the medical staff terminal 9, my data collection server 11 and the patient terminals 5-1, ..., (5-N).

The artificial intelligence learning unit 33 is executed every predetermined cycle (T) under the control of the control unit 30, and the medication compliance risk extraction algorithm used in the artificial intelligence-based medication compliance risk extraction unit 37 described below is learned. do.

At this time, the medication compliance risk extraction algorithm uses the effective information (input data 1) extracted from the patient's corpus and my data (input data 2) collected by the my data collection unit 36 as input data to output the medication compliance risk As described in detail in FIG. 4 , the medication compliance of the patient is output as a percentage and at the same time, medication compliance grades are output for each preset category.

In the learning of this medication compliance risk extraction algorithm, 'supervised learning' and 'unsupervised learning' proceed independently at the same time, and the characteristics of 'supervised learning' and 'unsupervised learning' of the present invention Is as follows.

- Supervised learning

The supervised learning of the medication compliance risk extraction algorithm analyzes input valid information (extracted from the patient corpus) and trains an extraction model for outputting medication compliance grades for each of preset categories.

In addition, the supervised learning of the medication compliance risk extraction algorithm uses the KNN (K-Nearest Neighbors) algorithm to classify and , In detail, when the input data is displayed on the coordinate plane, K data are analyzed to determine which label has the most, and labeling of the input data is performed.

For example, if a patient expresses distrust of the medical system, saying 'pharmaceutical companies are making bottles to make money', this classifies patients into a group with low trust in hospitals, medical staff, or pharmaceutical companies. This leads to a high risk of medication adherence.

That is, during supervised learning, the medication compliance risk extraction algorithm utilizes valid information (extracted from the patient's corpus) and preset labels to generate learning data capable of learning the correlation between valid information and labels, and the generated learning data. , derives an extraction model, which is a set of parameter values for correlation between valid information and labels, and verifies the extraction model by measuring accuracy by applying a test data set to the derived extraction model.

In other words, the supervised learning of the medication compliance risk extraction algorithm of the present invention sets a plurality of preset categories, divides the medication compliance grades (very good, normal, bad, and very bad) of each category into a plurality, and builds a model. By learning, it is possible to output medication compliance grades for each category of the patient through valid information extracted from the patient's corpus.

- Unsupervised learning

The unsupervised learning of the medication compliance risk extraction algorithm analyzes the input valid information (extracted from the patient corpus) and My Data, and trains an extraction model to output the patient's medication compliance rating for each of the preset categories.

In addition, the unsupervised learning of the medication compliance risk extraction algorithm uses a clustering technique, and automatically groups patients based on characteristics such as age, gender, medication period, adverse reaction, relationship with doctor, and my data, Its characteristics can be identified, and in detail, the number of clusters can be set and trained using the K-Means class in the Python Scikit-Learn package.

At this time, the characteristics such as age, gender, administration period, adverse reaction, and relationship with the doctor, which are used in unsupervised learning, are configured independently.

For example, when data on a new female patient in her 30s is registered, age, place of residence, medical history, lifestyle, attitude toward medication, reliability of the medical system, level of common sense in health and disease management, etc. along with financial transaction delinquency and Using My Data, such as card and communication fee payment status, it is possible to create a group that can be grouped similarly to the patient, and to identify the characteristics of the group to which the patient belongs. In detail, women in their 30s Among them, it is possible to derive implications such as a relatively high risk of medication compliance in the group of patients with card payment overdue.

In this way, the medication compliance risk extraction algorithm of the present invention simultaneously proceeds with supervised learning and unsupervised learning, and makes it possible to derive a more accurate and sophisticated model by complementing and utilizing the disadvantages of the models according to each method.

At this time, as the medication compliance risk extraction algorithm learned by the artificial intelligence learning unit 33, various models already known or developed in the future may be applied.

The medical team matching unit 34 receives basic information transmitted from the counseling application 7 through the communication interface unit 32 .

In addition, the medical staff matching unit 34 utilizes the type of disease included in the input basic information, information on medical facilities and medical staff who have performed previous medical treatment, or information on medical facilities and medical staff who wish to be treated, so that the doctor who will treat the patient match

At this time, the medical staff matching unit 34 matches the medical facility and medical staff desired by the patient included in the input basic information with the corresponding patient, or the medical facility and medical staff previously treated by the patient included in the input basic information to the corresponding patient. It can be operated in a way that matches with.

In addition, the matching information generated by the medical staff matching unit 34 is stored in the database unit 31 under the control of the control unit 30 .

8 is a block diagram illustrating an AI chatbot-based voice consultation service unit of FIG. 6 .

As shown in FIG. 8, the AI chatbot-based voice consultation service unit 35 includes a consultation request data input module 351, a data extraction module 352, a connection module 353, and an AI chatbot-based voice conversation module 354 ) is made up of

At this time, as described above, the voice counseling service unit 35 of the counseling service management server 3 can be made through a call with a professional counselor.

The consultation request data input module 351 receives consultation request data transmitted from the consultation application 7 through the communication interface unit 32 .

The data extraction module 352 searches the database unit 31 and extracts, from the database unit 31, the type of disease of the patient who has requested the consultation, and preset query words and keyword information corresponding to the type of disease.

The connection module 353 checks the connection with the consultation application 7 requested by the AI chatbot to the consultation service management server 3 for consultation.

The AI chatbot-based voice conversation module 354 performs AI-based voice counseling with a patient connected to the AI chatbot.

At this time, as described above, the AI chatbot performs a voice conversation with the patient connected based on artificial intelligence, but asks questions corresponding to each necessary check item according to the patient's disease type to obtain a conversation related to medication compliance extraction from the patient. induce

My data collection unit 36 is executed when voice consultation is started by the AI chatbot-based voice consultation service unit 35, and determines whether the patient agrees to collect my data or not, and if the corresponding patient If the patient agrees to collect My Data, the patient's identification information is transmitted to the My Data collection server 11 to request My Data, and then the My Data collection server 11 receives the patient's My Data Collect My Data with

9 is a block diagram illustrating an artificial intelligence-based medication compliance risk extraction unit of FIG. 6 .

The artificial intelligence-based medication compliance risk extraction unit 37 of FIG. 9 extracts effective information necessary for extracting the medication compliance risk from the patient's voice data, and then uses the medication compliance risk extraction algorithm learned in the above-described artificial intelligence learning unit 33. Based on the extracted effective information and the collected My Data, the risk of medication compliance of the patient is output.

At this time, valid information means meaningful information necessary for extracting the risk of medication compliance during conversation with the patient.

In addition, as shown in FIG. 9, the artificial intelligence-based medication compliance risk extraction unit 37 includes a voice data input module 371, a my data input module 372, a voice recognition and corpus generation module 373, and effective information It consists of an extraction module 374 and an AI-based medication compliance risk extraction module 375.

The voice data input module 371 receives voice data from the AI chatbot-based voice conversation module 354 of FIG. 8 described above.

The my data input module 372 receives the my data collected by the my data collection unit 36 .

The voice recognition and corpus generation module 373 analyzes the voice data input by the voice data input module 371, recognizes the voice, and generates a corpus of the recognized patient's voice data.

The valid information extraction module 374 analyzes the patient corpus generated by the speech recognition and corpus generation module 373 using a preset valid information extraction algorithm, and extracts valid information from the patient corpus. At this time, the valid information extraction algorithm is an algorithm for extracting valid information, which is meaningful information necessary for extracting the patient's medication compliance risk by analyzing the input corpus.

Also, although not shown in the figure, the valid information extraction module 374 includes a text preprocessing module and a feature vectorization module.

At this time, the text pre-processing module (not shown) performs pre-processing according to the purpose before converting unstructured text data into features, in detail, pre-processing such as tokenization, identity and normalization, stem extraction and headword extraction, stopword removal, and data separation. , and the feature vectorization module (not shown) converts the feature of the text preprocessed by the text preprocessing module into a vector value that is a numeric value.

In general, machine learning algorithms require feature vectorization because they operate by receiving numeric features as data. At this time, feature vectorization can use general-purpose algorithms such as Word2Vec, LDA, and FastText. 1) Word2Vec has a CBOW method and a Skip-Gram method, and both can be used in the present invention. As is widely known at this time, CBOW is a method of predicting words in the overall context, Skip-gram is a method of inferring surrounding words from words, and 2)LDA (Latent Dirichlet Allocation) is a process of finding topics in a set of documents. As a representative algorithm of topic modeling, it uses unsupervised learning and stochastic algorithms that find hidden topics from corpus and group topics by keyword.

Also, the valid information extraction module 374 extracts valid information from the patient corpus by applying a preset valid information extraction algorithm to the feature vectorized set.

At this time, the valid information extracted by the valid information extraction module 374 is input to the AI-based medication compliance risk output module 375 .

The AI-based medication compliance risk output module 375 receives valid information and my data from the valid information extraction module 374 and the my data input module 372.

In addition, the AI-based medication compliance risk output module 375 uses the input valid information and my data as input data of the medication compliance risk extraction algorithm learned by the artificial intelligence learning unit 33 to determine the medication compliance risk for the patient. print out

At this time, the AI-based medication compliance risk extraction algorithm outputs the medication compliance (percentage) of the patient and at the same time outputs medication compliance grades for each category.

The medication compliance report generating unit 38 utilizes the patient's medication compliance and medication compliance grades for each category extracted by the artificial intelligence-based medication compliance risk extraction unit 37 to provide a format of the medication compliance report 900 and After converting according to the type, a medication compliance report is created by filling in the corresponding blank.

At this time, when the medication compliance report is generated by the medication compliance report generating unit 38, the controller 30 transmits the medication compliance report generated to the medical staff terminal 9 matched to the patient by using matching information of the patient. The communication interface unit 32 is controlled so as to be.

On the other hand, as a method of providing the medication compliance report of the counseling service management server 3, various known methods such as encrypted e-mail, text message, dedicated program, and mail delivery may be applied.

10 is a flowchart for explaining an operation process of a medication compliance risk prediction system according to an embodiment of the present invention.

As shown in FIG. 10, the medication compliance risk prediction method (S1) of the present invention includes an artificial intelligence learning step (S10), a medical staff matching step (S20), a consultation request step (S30), and an AI chatbot connection step (S40). , AI chatbot-based voice counseling step (S50), my data collection step (S60), effective information extraction step (S70), artificial intelligence-based medication compliance risk extraction step (S80), medication compliance report generation step (S90), medication compliance It consists of a report transmission step (S100).

In the artificial intelligence learning step (S10), the counseling service management server 3 uses the effective information and my data extracted from the patient corpus as input data, and the medication compliance risk extraction algorithm, which is a machine learning algorithm for outputting the patient's medication compliance risk It is a step of learning every preset period (T).

At this time, since the learning technology and method of the artificial intelligence learning step (S10) have been described in the artificial intelligence learning unit 33 of FIG. 6 described above, a detailed description thereof will be omitted.

In addition, the medication compliance risk extraction algorithm learned by the artificial intelligence learning step (S10) is used in the artificial intelligence-based medication compliance risk extraction step (S80).

In the medical staff matching step (S20), when the consultation service management server 3 receives basic information from the consultation application 7, it matches a doctor who will treat the corresponding patient by using the received basic information.

Also, in the medical staff matching step (S20), matching information of the matched patient and medical staff terminal is stored in the database unit 31.

Consultation request step (S30) is a step in which the consultation application 7 transmits consultation request data to the consultation service management server 3 according to the request of the user (patient or patient guardian).

The AI chatbot connection step (S40) is a step in which the AI chatbot of the consultation service management server 3 connects to the corresponding consultation application 7.

The AI chatbot-based voice counseling step (S50) proceeds after the AI chatbot connection step (S40), and is a step in which the AI chatbot proceeds with the patient and artificial intelligence-based voice counseling service through the counseling application (7).

My data collection step (S60) is a step of collecting my data related to the patient, which proceeds after the AI chatbot access step (S40).

The effective information extraction step (S70) analyzes the voice data input from the AI chatbot-based voice counseling step (S50) to recognize the voice, and then uses the detected recognition information to generate a corpus of the patient, and the generated After preprocessing the unstructured text data of the corpus before converting it into features, the features of the preprocessed text are converted into vector values, which are numeric values, and a valid information extraction algorithm is applied to the feature vectorized set, valid information from the patient corpus. is the step of extracting

At this time, valid information means meaningful information necessary for extracting the risk of medication compliance during conversation with the patient.

The artificial intelligence-based medication compliance risk extraction step (S80) receives the valid information extracted by the valid information extraction step (S70) and my data collected by the my data collection step (S60), the input valid information and my data Using the data as input data of the medication compliance risk extraction algorithm learned by the artificial intelligence learning step (S10), the medication compliance risk of the corresponding patient is output. At this time, the output risk of medication compliance includes medication compliance (percentage) of the corresponding patient and medication compliance grade information for each category.

In the medication compliance report generation step (S90), the medication compliance report output by the counseling service management server 3 in the artificial intelligence-based medication compliance risk extraction step (S80) and the medication compliance grade information of each category are stored in a preset medication compliance report format. and generating a medication compliance report by converting according to the type and filling in the corresponding blank.

In the medication compliance report transmission step (S100), the medication compliance report generated by the medication compliance report generation step (S90) by the counseling service management server 3 is sent to the patient's counseling application 7 and the medical staff terminal matched to the patient ( This is the step of sending to 9).

Hereinafter, a second embodiment of the counseling service management server of the present invention will be described.

In general, there are cases of patients who give necessary answers to the doctor's questions during medical consultation, but in the case of patients with related diseases accompanied by deterioration of cognitive function, such as old age or mental health medicine, their basic comprehension and expression skills are relatively low. Therefore, a problem arises in that meaningless conversation time (hereinafter referred to as consumption time t2) increases in the medical consultation time for reasons such as having free conversation with a doctor's question or having a conversation at a very slow speed.

Table 3 below is an example of a conversation pattern according to the patient's comprehension and expressive power during medical consultation with a patient with normal cognitive decline.

[Table 3]

Referring to Table 3, looking at the conversation patterns of patients with cognitive decline, in the case of patients with cognitive decline, the consumption time (t2) increased compared to general patients, regardless of high or low comprehension and expression. , In particular, in the case of patients with low understanding or expressive power, it can be seen that the consumption time (t2) increases remarkably.

FIG. 11 is a block diagram illustrating a second embodiment of a counseling service management server of FIG. 6 .

The second consultation service management server 23 of FIG. 11 is the second embodiment of the consultation service management server of the present invention, and is intended to increase consultation efficiency by reducing the consumption time t2 described above during voice consultation of the AI chatbot .

In addition, as shown in FIG. 11, the second counseling service management server 23 includes a communication interface unit 32, an artificial intelligence learning unit 33, and a medical team matching unit 34 having the same configuration and operation as those of FIG. 6 described above. ), a my data collection unit 36, an AI-based medication compliance risk extraction unit 37, a medication compliance report generation unit 38, and an additional service unit 39.

In addition, the second counseling service management server 23 includes a second control unit 230, a second database unit 231, a second AI chatbot-based voice consultation service unit 235, an output determination unit 237, A counseling level setting unit 238 and a ventilation comment output period T1 detection unit 239 are further included.

The second control unit 230 performs the same operation as the control unit 30 of FIG. 6 described above, but when the voice consultation starts in the second AI chatbot-based voice consultation service unit 235, the output decision unit 237 ) is executed, but when the voice counseling ends, the counseling level setting unit 238 is executed.

In addition, the second control unit 230 executes the detection unit 239 for the output period of ventilation comments (T1) when the consultation level of the patient is set by the consultation level setting unit 238.

The second database unit 231 stores the same data as the database unit 31 of FIG. 6 described above, but stores a reference table in which consultation grades L are matched for each consumption rate R section. At this time, the consumption rate (R) means the percentage occupied by questions and free conversations during the patient's voice counseling, and will be described in detail with reference to FIGS. 13 and 14 to be described later.

In addition, the second database unit 231 stores a second reference table in which ventilation comment output cycles T1 are matched for each consultation level L. At this time, the ventilation comment means a repeated comment for the AI chatbot to naturally interrupt the patient's conversation and ask the question again, and the ventilation comment output period (T1) means the maximum value of the continuous consumption time (t2).

In addition, the second database unit 231 stores the consultation level information of the corresponding patient set in the consultation level setting unit 238 .

In addition, when it is assumed that the higher the counseling grade (L) is the higher level, the ventilation period output period (T1) of the second reference table is set to be proportional to the counseling grade (L). That is, the ventilation period output period T1 of the second reference table is set longer as the consultation level (L) is higher, and is set shorter as the consultation level (L) is lower.

For example, assuming that the higher the counseling grade (L) is the higher level, and the levels of the counseling grade (L) are classified from '1' to '10', having an excellent counseling grade (L) of level 10 The ventilation comment output period T1 of the patient A may be set long, and the ventilation comment output period T1 of the patient B having a low counseling level L of level 1 may be set short.

In general, patients with cognitive decline, such as old age and mental illness, have a characteristic that the consumption time (t2), which is meaningless content, increases during voice counseling, and this characteristic not only reduces the efficiency of counseling but also reduces the overall conversation Due to the increase in time, unnecessary calculation throughput of processors increases. The present invention is intended to solve this problem, and is configured to detect the counseling grade calculated based on the patient's previous voice counseling and the output cycle (T1) of the ventilation comment according to the counseling grade (L), based on the second AI chatbot The voice counseling service unit 235 is configured to output an arousal when the patient's consumption time (t2) continues, so that the patient's concentration is naturally aroused to induce an answer to the desired question, as well as to reduce the consumption time (t2). can be significantly reduced.

The second AI chatbot-based voice consultation service unit 235 performs the same operation as the AI chatbot-based voice consultation service unit 35 of FIG. 6 described above, but is given to the patient by the consultation level setting unit 238 described later Depending on the level (L) of the consultation, the difficulty or speed of the conversation may be different.

For example, when the second AI chatbot-based voice counseling service unit 235 receives an output control signal from the output control signal from the output control unit 237 of FIG. .

At this time, since the length of the ventilation comment output period (T1) is determined according to the counseling level (L) according to the patient's previous counseling history, the patient's consumption time (t2) can be remarkably reduced.

FIG. 12 is a block diagram illustrating a determination unit for outputting a ventilation message in FIG. 11 .

12 is executed when the voice consultation of the AI chatbot starts through the second AI chatbot-based voice consultation service unit 235.

In addition, as shown in FIG. 12, the ventilation message output determination unit 237 includes a data search and extraction module 2371, a monitoring module 2372, an elapsed time (Δt) calculation module 2373, and a comparison module ( 2374), and a ventilation moment output determination module 2375.

The data search and extraction module 2371 searches the second database unit 231 and searches for a ventilation comment output period T1 for the corresponding patient.

In addition, the data search and extraction module 2371 1) when the ventilation comment output period T1 for the patient is searched for, extracts the ventilation comment output period T1, 2) the ventilation comment output period for the patient ( T1) is not searched, and the corresponding patient is determined as a new patient and no additional operation is performed thereafter.

The monitoring module 2372 is executed when the ventilation comment output period T1 for the patient is extracted by the data search and extraction module 2371, and the second AI chatbot-based voice counseling service unit 235 and the above-described figure The effective information extraction module 374 of 9 is monitored.

When monitored by the monitoring module 2372, the elapsed time (Δt) calculation module 2373 calculates the elapsed time (Δt) from the time when the AI chatbot asks the query to the present in which valid information is not extracted. yield

The comparison module 2374 compares the elapsed time (Δt) calculated by the elapsed time (Δt) calculation module 2373 and the excitation output period (T1) extracted by the data search and extraction module 2371. In detail, whether the elapsed time (Δt) reaches the ventilation moment output period (T1) is compared (Δt≥T1).

In the comparison module 2374, the ventilation moment output determination module 2375 determines that: 1) If the elapsed time (Δt) is less than the ventilation moment output period (T1) (Δt<T1), the AI chatbot has not output the ventilation comment yet 2) If the elapsed time (Δt) is greater than or equal to (Δt≥T1) the ventilation comment output period (T1), it is determined that the AI chatbot should output the ventilation comment.

In addition, when the output decision module 2375 determines the output of the AI chatbot, it outputs an output control signal to the second AI chatbot-based voice consultation service unit 235 .

At this time, the second AI chatbot-based voice counseling service unit 235 outputs preset voice data of the excitement message when an output control signal is input from the decision unit 237 whether to output the excitement message.

FIG. 13 is a block diagram showing the counseling level setting unit of FIG. 11 , and FIG. 14 is an exemplary diagram for explaining the total output time (t), effective time (t1), and consumption time (t2) of FIG. 13 .

As shown in FIG. 13, the equivalent level setting unit 238 includes a data input module 2381, a voice data classification module 2382, a total output time (t) calculation module 2383, and valid time (t1) calculation It consists of a module 2384, a consumption time (t2) calculation module 2385, a consumption rate (R) calculation module 2386, and a consultation level determination module 2387.

The data input module 2381 receives the voice data of the voice counseling performed by the second AI chatbot-based voice counseling service unit 235 and the valid information extracted by the valid information extraction module 374 of FIG. 9 described above. .

The voice data classification module 2382 sorts the input voice data into a bit string according to the timeline, and references and utilizes the input valid information to validate the voice data arranged in the bit string as shown in FIG. 14. It is classified into a section in which information is output and a section in which valid information is not output.

The total output time (t) calculation module 2383 calculates the total output time (t), which is the total output time of voice data.

The valid time t1 calculation module 2384 calculates the valid time t1, which is an output time obtained by summing the intervals for which valid information is output by the voice data classification module 2382.

The consumption time t2 calculation module 2385 calculates the consumption time t2, which is an output time obtained by summing the sections in which valid information is not output by the voice data classification module 2382.

The consumption rate (R) calculation module 2386 calculates the consumption time (t2) calculated by the consumption time (t2) calculation module 2385 and the total output time (t) calculated by the total output time (t) calculation module 2383 Calculate the consumption rate (R), which is a percentage of ).

That is, during voice counseling, the consumption rate (R) increases as the patient engages in unnecessary conversation unrelated to the question, and the consumption rate (R) decreases as the patient answers necessary questions to the question.

The consultation level determination module 2387 searches the reference table stored in the database unit 31 and detects a consultation level L corresponding to the consumption rate R calculated by the consumption rate R calculation module 2386; The detected consultation level (L) is determined as the consultation level (L) of the patient.

At this time, the reference table is data in which the counseling level (L) is matched for each section of the consumption rate (R).

In addition, the consultation level (L) determined by the consultation level determination module 2387 is stored in the database unit 31 under the control of the control unit 30 .

Returning to FIG. 11 and examining the ventilation comment output period (T1) detection unit 239, the ventilation comment output period (T1) detection unit 239 determines the counseling level (L) of the corresponding patient set by the consultation level setting unit 238. Upon receiving the input, the second reference table stored in the second database unit 231 is referred to and utilized to detect the ventilation comment output period T1 matching the consultation level L of the corresponding patient.

At this time, as described above, when the consultation level (L) is higher, the length of the output period (T1) is set longer as the consultation level (L) is higher and shorter as the consultation level (L) is lower, as described above.

Fig. 15 is a flowchart showing the second embodiment of Fig. 10;

The second medication compliance risk prediction method (S200) of FIG. 15 includes artificial intelligence learning step (S210), medical staff matching step (S220), consultation request step (S230), AI chatbot access step (S240), AI chatbot-based voice counseling Step (S250), my data collection step (S260), effective information extraction step (S270), artificial intelligence-based medication compliance risk extraction step (S280), monitoring step (S290), voice counseling end determination step (S300), ventilation Comment output determination step (S310), output control signal output step (S320), medication compliance report generation step (S330), consultation level setting step (S340), ventilation comment output cycle (T1) detection step (S350), medication compliance It consists of a report transmission step (S360).

At this time, the artificial intelligence learning step (S210) of FIG. 15, medical staff matching step (S220), consultation request step (S230), AI chatbot connection step (S240), my data collection step (S260), valid information extraction step (S270), The artificial intelligence-based medication compliance risk extraction step (S280), medication compliance report generation step (S330), and medication compliance report transmission step (S360) are the artificial intelligence learning step (S10) of FIG. 10, medical staff matching step (S20), Consultation request step (S30), AI chatbot access step (S40), my data collection step (S60), valid information extraction step (S70), artificial intelligence-based medication compliance risk extraction step (S80), medication compliance report generation step (S90) ) And the medication compliance report transmission step (S100) proceeds in the same process, so a detailed description will be omitted.

The AI chatbot-based voice consultation step (S250) proceeds after the AI chatbot and counseling application 7 are connected by the AI chatbot connection step (S240), and the AI chatbot-based voice consultation step (S250) of FIG. 10 and The same process is performed, but when the output control signal is transmitted from the output control signal output step 320 to be described later, the AI chatbot outputs the preset voice data of the excitement.

The monitoring step (S290) is a step of monitoring the AI chatbot-based voice consultation step (S250) and the artificial intelligence-based effective information extraction step (S270).

In addition, in the monitoring step (S290), the second database unit 231 is searched for, and a ventilation comment output period (T1) for the corresponding patient is searched.

In addition, the monitoring step (S290) calculates the elapsed time (Δt) from the time when the AI chatbot uttered a query in the AI chatbot-based voice counseling step (S250) to the current state in which valid information is not extracted.

The voice consultation end determination step (S300) is a step of determining whether the voice consultation has ended in the AI chatbot-based voice consultation step (S250) through the monitoring step (S290).

In addition, the step of determining whether voice counseling is terminated (S300) is 1) if the voice counseling of the AI chatbot ends, the next step is to generate a medication compliance report (S330) and set a counseling level (S340), and 2) if AI If the voice counseling of the chatbot is in progress, the next step is to determine whether to output an inspirational message (S310).

The step of determining whether to output an arousal message (S310) is performed when it is determined that the voice counseling of the AI chatbot is in progress in the step of determining whether to end voice counseling (S280), and the elapsed time (Δt) calculated in the monitoring step (S270) and the corresponding Compare the ventilation ment output cycle (T1) of the patient.

In addition, in the step of determining whether to output the ventilation message (S290), 1) if the elapsed time (Δt) is less than the ventilation message output period (T1) (Δt<T1), proceed with the monitoring step (S290) as the next step, 2) If the elapsed time (Δt) is longer than the ventilation moment output period (T1) (Δt≥T1), the output control signal output step (S320) proceeds to the next step.

The output control signal output step (S320) proceeds when the elapsed time (Δt) in the ventilation comment output determination step (S310) is greater than or equal to the ventilation comment output period (T1) (Δt≥T1), and the output is controlled by AI chatbot This step is to output a signal.

The counseling level setting step (S340) is performed when it is determined that the AI chatbot's voice counseling has ended in the voice counseling end determination step (S300), and the voice data of the voice counseling conducted in the AI chatbot-based voice counseling step (S250), After calculating the total output time (t), valid time (t1), and consumption time (t2) of the patient's voice data by using the valid information extracted by the artificial intelligence-based valid information extraction step (S270), Calculate the consumption rate (R).

In addition, in the consultation level setting step (S340), the consultation level (L) of the patient is determined using the calculated consumption rate (R) and the reference table, and the determined consultation level (L) is stored in the database unit 31.

Arousing message output cycle (T1) detection step (S350) proceeds after consultation level setting step (S340), consultation service management server 3 searches for a preset second standard table, consultation level determination step (S340) This is a step of detecting a ventilation moment output period T1 matched to the counseling level L of the patient determined by , and storing the detected ventilation moment output period T1 in a database unit.

As described above, the medication compliance risk prediction system 1, which is an embodiment of the present invention, provides a voice counseling service to a patient or guardian, extracts effective information, which is meaningful information necessary for extracting the patient's medication compliance risk, from the voice counseling, and learns Using a machine learning algorithm, the effective information extracted and collected my data (financial transactions, communication expenses, etc.) are analyzed to output the patient's medication compliance risk, and a medication compliance report is generated based on the output medication compliance risk information. By providing this to the medical staff and the insurance company server, the medical staff performs medical consultation by referring to the medication compliance report, so that the patient's medication compliance can be dramatically increased, and medical costs due to conventional low medication compliance can be reduced. , has the advantage of maximizing counseling concentration, efficiency and effect compared to the same consultation time.

In addition, the medication compliance risk prediction system 1 of the present invention can reduce the patient's recurrence or deterioration as the counseling service management server provides voice counseling service with the patient, thereby significantly reducing the patient's treatment cost and saving the patient's time. And medical consultation is possible without space limitation, so the convenience of medical treatment can be improved.

In addition, in the medication compliance risk prediction system 1 of the present invention, after the counseling service management server collects the patient's My Data, the collected My Data is used as input data for the medication compliance risk extraction algorithm to extract the patient's medication compliance risk. accuracy and reliability can be further improved.

In addition, the medication compliance risk prediction system 1 of the present invention simultaneously and individually proceeds with supervised learning and unsupervised learning when the medication compliance risk extraction algorithm learns, so the disadvantages of the model according to each method By being configured to complement each other, it is possible to extract the risk of medication compliance more precisely.

In addition, in the medication compliance risk prediction system 1 of the present invention, the counseling level setting unit calculates the consumption time t2 consisting of meaningless information among the patient's voice data, and then the patient's consultation level according to the calculated consumption time t2. (L) is detected, and the AI chatbot proceeds with different difficulty or speed of conversation for each consultation level (L), thereby further increasing the efficiency and effectiveness of medical consultation.

In addition, the medication compliance risk prediction system 1 of the present invention refers to the medication compliance report transmitted by the insurance company server from the counseling service management server, and differentiates insurance premiums according to the risk of medication compliance at the time of insurance subscription or when insurance money is compensated. By paying insurance money differently according to the risk of compliance, unnecessary budget consumption by insurance companies can be drastically reduced, and the disease treatment rate can be increased by increasing the patient's willingness and participation rate in improving medication compliance.

1: Medication compliance risk prediction system 3: Consultation service management server
5-1, ..., 5-N: patient side terminals 7: counseling application
9: medical staff terminal 10: communication network
11: My data collection server 30: control unit
31: database unit 32: communication interface unit
33: artificial intelligence learning unit 34: medical team matching unit
35: AI chatbot-based voice counseling service department 36: My data collection department
37: Artificial intelligence-based medication compliance risk extraction unit
38: Medication compliance report generation unit 39 Additional service unit
51: input means 52: audio output means
53: voice collection means 54: display means
55: communication means 351: consultation request data input module
352: data extraction module 353: connection module
354: AI chatbot-based voice conversation module 371: Voice data input module
372: My data input module 373: Voice recognition and corpus generation module
374: Effective information extraction module 375: AI-based medication compliance risk output module

Claims (11)

Medication compliance risk prediction system including patient terminal and counseling service management server:
The consultation service management server
Check items necessary for extracting the patient's medication compliance risk, queries for inducing answers to each check item, and keywords that can be determined as words related to each check item are preset and stored, and the patient's medication intake is stored. a database unit storing an effective information extraction algorithm for extracting valid information, which is meaningful information necessary for compliance risk extraction, and a medication compliance risk extraction algorithm for outputting a medication compliance risk by analyzing the extracted valid information;
It is executed every predetermined period (T), generates learning data capable of learning the correlation between the effective information and the risk of medication compliance, and utilizes the generated learning data to obtain parameter values for the correlation between the effective information and the risk of medication compliance. an artificial intelligence learning unit that performs learning of the medication compliance risk extraction algorithm by deriving an extraction model which is a set;
a voice counseling service unit that connects to the patient-side terminal and executes voice counseling at the request of the patient-side terminal;
After recognizing the patient's voice data input from the voice counseling service unit, generating a corpus, valid information is extracted from the generated patient corpus using the valid information extraction algorithm, and the extracted valid information is an artificial intelligence-based medication adherence risk extraction unit that outputs a medication compliance risk of the patient based on input data of the medication compliance risk extraction algorithm learned by the artificial intelligence learning unit;
a medication compliance report generation unit generating a medication compliance report based on the medication compliance risk information extracted by the artificial intelligence-based medication compliance risk extraction unit;
It is executed when the voice counseling service ends, and when the patient's voice data executed through the voice counseling service unit is input, the inputted patient's voice data is analyzed according to a preset rating procedure to communicate with the patient. After setting a consultation level (L) according to the ability level, a consultation level setting unit for storing in the database unit;
a control unit for transmitting the medication compliance report generated by the medication compliance report generation unit to a medical staff terminal of a corresponding patient;
The voice counseling service department
When the counseling level (L) of the patient requesting consultation is stored in the database unit, the difficulty or speed of the conversation is different for each consultation level (L) of the patient,
The consultation level setting unit
a data input module that receives the patient's voice data executed through the voice counseling service unit and valid information extracted by the artificial intelligence-based medication compliance risk extraction unit;
Voice data input through the data input module is sorted into a bit string according to the timeline, and the input valid information is referenced and utilized to output the valid information output section and valid information for the voice data arranged in the bit string. a voice data classification module that classifies the voice data into non-intervals;
a total output time (t) calculation module that calculates a total output time (t), which is a total output time of the audio data input by the data input module;
a consumption time (t2) calculating module that calculates a consumption time (t2), which is an output time obtained by adding up sections in which valid information is not output by the voice data classification module;
Consumption rate (R) that calculates the consumption rate (R), which is a percentage of the consumption time (t2) calculated by the consumption time (t2) calculation module and the total output time (t) calculated by the total output time (t) calculation module calculation module;
Using the reference table in which the counseling grade (L) is matched for each section of the consumption rate (R), the counseling grade (L) corresponding to the consumption rate (R) calculated by the consumption rate (R) calculation module is detected, and the detected After determining the counseling grade (L) as the counseling grade (L) of the patient, the medication compliance risk prediction system comprising a counseling grade determination module for storing in the database unit. The method of claim 1, wherein the medication compliance risk prediction system
For patients requested from the above counseling service management server, past health examination results, account information, loan payment information, automobile insurance contract and payment information, insurance loan product information, card use information, communication fee payment information, guarantee insurance information, national tax payment Further comprising a my data collection server for collecting my data including at least one or more of information, local tax payment information, and four major insurance payment information,
The consultation service management server
It is executed when voice counseling is started by the voice counseling service unit, and after requesting My Data for the corresponding patient from the My Data collection server, the My Data collection unit receives My Data of the corresponding patient from the My Data collection server. include,
The medication compliance risk extraction algorithm
Taking the effective information extracted by the artificial intelligence-based medication compliance risk extraction unit and the my data collected by the my data collection server as input data, the medication compliance risk prediction system, characterized in that for outputting. The method of claim 2, wherein the artificial intelligence learning unit simultaneously and separately proceeds with supervised learning and unsupervised learning when learning the medication compliance risk extraction algorithm, thereby eliminating the disadvantages of the model according to each method. Medication compliance risk prediction system, characterized in that to complement each other. The method of claim 3, wherein the artificial intelligence-based medication compliance risk extraction unit
a voice data input module that receives the patient's voice data from the voice counseling service unit;
My data input module receiving my data from the my data collection unit;
a voice recognition and corpus generating module that recognizes voice by analyzing the voice data input by the voice data input module and generates a corpus of the recognized patient's voice data;
a valid information extraction module for extracting valid information from the patient corpus by analyzing the patient corpus generated by the voice recognition and corpus generating module using the valid information extraction algorithm;
When valid information and my data are received from the valid information extraction module and the my data input module, the input valid information and my data are used as input data of the learned medication compliance risk extraction algorithm to determine the medication compliance risk for the patient A medication compliance risk prediction system comprising an AI-based medication compliance risk output module for outputting. The method of claim 4, wherein the medication compliance risk prediction system further comprises an insurance company server,
The insurance company server
Pre-designed insurance products are provided, and when a patient subscribes to an insurance product or pays compensation, report request data including the patient's identification information is transmitted to the counseling service management server, and medication compliance received from the counseling service management server Depending on the risk of medication adherence in the report, premiums for the insurance products subscribed to are differentiated or compensation is paid differently.
The medication compliance risk prediction system, characterized in that, when the counseling service management server receives report request data from the insurance company server, it transmits a medication compliance report of the corresponding patient to the insurance company server. The method of claim 4, wherein the medication compliance risk output by the medication compliance risk extraction algorithm includes medication compliance of the corresponding patient and medication compliance grades for each category,
The medication compliance report
A medication compliance risk prediction system, characterized in that for the patient, the medication compliance is displayed as a percentage and at the same time, medication compliance grades of each category are displayed. The method of claim 4, wherein a counseling application is installed in the patient-side terminal,
The counseling application
When receiving a consultation request from a user (patient), the consultation request data is transmitted to the consultation service management server and a graphical user interface (GUI) is provided to provide an interface for the user (patient) and voice consultation service of the consultation service management server. , Graphic User Interface), a medication compliance risk prediction system, characterized in that for displaying an interface for counseling on the monitor of the patient-side terminal. The method of claim 7, wherein the counseling application
When the user (patient) receives basic information including personal information, disease type, previous medical staff information, and desired medical staff information, the input basic information is transmitted to the consultation service management server. do,
The consultation service management server
When basic information is received from the consultation application, the received basic information is stored in the database and at the same time, a medical staff who has performed previous medical treatment included in the basic information or a medical staff who wants to be treated is matched with a medical staff who will treat the patient. Medication compliance risk prediction system, characterized in that it further comprises a matching unit. delete delete The method of claim 4, wherein the database unit further stores a second reference table in which ventilation comment output cycles (T1) are matched for each consultation level (L),
The ventilation cycle output period (T1) of the second reference table is set to be proportional to (or inversely proportional to) the consultation level (L), assuming that the higher the consultation level (L), the higher the level (or lower level),
The consultation service management server
It is executed when the counseling grade is determined by the counseling grade setting unit, and the ventilation comment output period (T1) matching the counseling grade (L) of the patient determined by the consultation level setting unit is set using the second reference table. After detecting, a ventilation moment output period (T1) detection unit for storing in the database unit;
It is executed when the voice counseling by the voice counseling service unit starts, and further includes an output determination unit that terminates when the voice counseling ends,
The ventilation moment output determination unit
When first executed, a data search and extraction module for searching the database unit and searching for a ventilation comment output period (T1) for the patient;
a monitoring module that is executed when the output cycle (T1) of the patient is extracted by the data search and extraction module, and monitors the voice counseling service unit and the effective information extraction module;
An elapsed time (Δt) calculation module for calculating an elapsed time (Δt) from the time when the voice counseling service unit asks the query to the present in which a state in which valid information is not extracted continues, when monitored by the monitoring module;
a comparison module for comparing the elapsed time (Δt) calculated by the elapsed time (Δt) calculation module with the excitation output period (T1) extracted by the data search and extraction module;
In the comparison module, if the elapsed time (Δt) is more than the ventilation comment output period (T1) (Δt≥T1), the ventilation comment is reached (Δt < T1), after determining that the AI chatbot should output the ventilation comment , A ventilation message output determining module for outputting an output control signal to the voice counseling service unit,
The voice counseling service department
When an output control signal is input from the ventilation comment output determination unit, the medication compliance risk prediction system, characterized in that for outputting a ventilation comment meaning a repetitive comment for interrupting the patient's conversation and asking the previous query again.

Images