AI-based virtual third-party neutral evaluation and mediation methods and systems for medical disputes

Abstract

This invention relates to the field of medical dispute resolution and discloses an AI-powered virtual third-party neutral evaluation and mediation method and system for medical disputes. The AI-powered virtual third-party neutral evaluation and mediation method includes: accepting case details, treatment process, and the demands of both parties from medical institutions; initiating privacy data desensitization verification to remove all personal identity information; calling upon a preset virtual third-party neutral evaluation role to conduct an objective comparative analysis of the treatment behaviors involved in the dispute, in conjunction with clinical treatment guidelines; extracting the core points of contention between the doctor and patient, transforming professional medical knowledge into easily understandable content, and simultaneously generating multiple versions of on-site prepared statements; generating a standardized evaluation opinion with a compliance disclaimer for use by medical institutions in on-site mediation; and archiving the dispute resolution process without privacy protection to complete the closed-loop dispute resolution process. This invention has the advantage of improving the efficiency of dispute resolution.

Description

Technical Field

[0001] This invention relates to the field of medical dispute resolution, specifically to an AI-based virtual third-party neutral evaluation and mediation method and system for medical disputes. Background Technology

[0002] With the rapid popularization of internet healthcare, the medical service process is gradually evolving from the traditional offline face-to-face model to a collaborative model involving multiple entities, platforms, and data sources. Disputes between doctors and patients regarding treatment, liability allocation, cost disputes, and information interpretation are becoming increasingly complex, unstructured, and frequent. Current methods for handling doctor-patient disputes typically rely on manual mediation, administrative coordination, or judicial appraisal. These methods heavily depend on expert experience, manual review of medical records, and multiple rounds of communication and confirmation. This not only results in long processing times and high mediation costs, but also, when dealing with electronic medical records, remote consultation records, AI-assisted diagnostic results, and multi-source medical behavior logs, difficulties in establishing evidence chains, inconsistent liability analyses, and highly subjective mediation conclusions can arise. Especially in scenarios where AI is involved in treatment recommendations, assisted image analysis, or automatic risk warnings, traditional dispute resolution mechanisms struggle to uniformly analyze the correlation between AI output behavior, doctor adoption behavior, and patient informed consent, leading to blurred boundaries of liability.

[0003] Most existing intelligent mediation systems only operate at the level of rule retrieval, case matching, or simple question-and-answer, lacking a dynamic and neutral evaluation mechanism covering the entire process of medical disputes. They are unable to form a continuous reasoning structure to address temporal changes, deviations in treatment decisions, and multi-party interactions in medical practice. When the medical process involves multi-departmental collaboration, multiple adjustments to treatment plans, rapid changes in the patient's condition, or deviations between AI suggestions and human decisions, existing systems struggle to identify key turning points of responsibility, easily leading to one-sided mediation opinions, inaccurate extraction of points of contention, and insufficient credibility of the mediation.

[0004] Therefore, it is necessary to design an AI-powered virtual third-party neutral review and mediation method and system for medical disputes to improve the efficiency of dispute resolution. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an AI-based virtual third-party neutral review and mediation method and system for medical disputes, which has the advantage of improving dispute resolution efficiency and solves the problems mentioned in the background technology.

[0006] To achieve the aforementioned goal of improving dispute resolution efficiency, this invention provides the following technical solution: an AI virtual third-party neutral review and mediation system for medical disputes, comprising:

[0007] Data anonymization module: Receives medical dispute materials entered by the hospital, performs structured parsing of medical record texts, examination reports, treatment records, doctor-patient communication content and auxiliary examination data, automatically identifies and masks patient identity information, contact information and related privacy fields, and completes data anonymization verification;

[0008] The diagnostic and treatment standard comparison module connects to a general clinical diagnostic and treatment knowledge base, objectively comparing the patient reception behavior, examination behavior, treatment behavior and disclosure behavior involved in the dispute with the industry's routine diagnostic and treatment standards, and identifying behavioral differences and missing information nodes in the diagnostic and treatment process;

[0009] The Dispute Analysis Module extracts the core demands and conflicts between doctors and patients, conducts correlation analysis on frequently disputed statements, emotionally conflicting statements, and content related to cognitive biases in diagnosis and treatment in dispute materials, and transforms professional medical content into popular interpretations.

[0010] Mediation script generation module: Based on the type of dispute, the intensity of the controversy, and the emotional state of the doctor and patient, it automatically generates three types of on-site mediation scripts: empathic version, professional version, and popular version. It also dynamically adjusts the order of explanation, intensity of expression, and reassurance strategies based on communication feedback.

[0011] The evaluation document output module integrates the results of comparison with treatment guidelines, analysis of points of contention, and mediation and communication results in a structured manner to generate a standardized neutral evaluation opinion. The document includes a built-in compliance disclaimer and supports three commercial adaptation modes: single case service, annual package service, and regional centralized procurement service.

[0012] Preferably, the process of completing data anonymization and verification is as follows:

[0013] The system performs structured parsing on medical records, examination reports, treatment records, doctor-patient communication information, and supporting materials entered by the hospital, and extracts the corresponding identity fields, time-related fields, and treatment-related fields.

[0014] Based on the entity recognition results, tiered masking is applied to the patient's name, ID number, contact number, home address, and hospitalization number;

[0015] Establish cross-mapping relationships for recurring related identity information in different medical records, and perform unified replacement processing on the mapped identity fields;

[0016] Semantic association scanning is used to identify implicit descriptions that reveal patient identity, and these implicit descriptions are then reconstructed by blurring.

[0017] By combining the proportion of fields remaining in the anonymized data, the degree of semantic integrity, and the consistency of information association, the anonymization results are validated and analyzed to form data anonymization results that meet privacy protection requirements.

[0018] Preferably, the process of identifying behavioral discrepancies and missing information nodes in the diagnosis and treatment process is as follows:

[0019] Organize the patient reception records, examination records, treatment records, and communication records involved in the dispute into a time sequence to form a corresponding chain of medical behavior;

[0020] Each behavioral node in the diagnosis and treatment behavior chain is matched with the standard diagnosis and treatment path in the clinical diagnosis and treatment knowledge base to identify diagnosis and treatment behaviors with sequence deviations, missing content, or abnormal execution intervals.

[0021] Based on the time span, operational continuity, and diagnostic dependence among behavioral nodes, abnormal behavioral fragments are clustered to form corresponding behavioral difference nodes;

[0022] Perform a completeness scan on the information provided, signed records, and examination feedback results in the diagnosis and treatment process chain to identify unrecorded, unfeasible, or interrupted process areas and form corresponding missing information nodes.

[0023] The preferred method for transforming professional medical content into accessible explanations is as follows:

[0024] Semantic segmentation is performed on the medical terminology, descriptions of diagnosis and treatment, and explanations of examination results output by the dispute resolution module to extract the professional expression fragments;

[0025] Based on the preset medical terminology mapping rules, the clinical meaning, treatment purpose, and risk description of professional expression fragments are replaced with popular semantics.

[0026] Based on the emotional fluctuations and cognitive differences between doctors and patients during the communication process, the expression level and sentence complexity of the simplified explanations are dynamically adjusted.

[0027] To address the descriptions of medical procedures that are prone to misunderstanding, a scenario-based explanatory structure is introduced to provide a related explanation of the reasons for diagnosis and treatment, the process, and the results.

[0028] The semantic consistency check is used to logically correct the converted interpretation, resulting in a simplified interpretation suitable for doctor-patient mediation scenarios.

[0029] This invention also discloses another technical solution: an AI-powered virtual third-party neutral review and mediation method for medical disputes, comprising the following steps:

[0030] Upon receiving information from medical institutions regarding medical disputes, treatment processes, and the demands of both parties, a privacy data anonymization verification process is initiated to remove all personally identifiable information.

[0031] By invoking a pre-defined virtual third-party neutral review role and combining it with clinical treatment guidelines, an objective comparative analysis of the treatment behaviors involved in the dispute is conducted.

[0032] Extract the core points of contention between doctors and patients, transform professional medical knowledge into popular interpretations, and simultaneously generate multiple versions of on-site scripts for adjustment.

[0033] Generate standardized review reports with compliance disclaimers for use by medical institutions in on-site mediation;

[0034] The dispute resolution process is archived without privacy restrictions to complete the closed-loop dispute resolution process.

[0035] The preferred method for conducting an objective comparative analysis of the medical practices involved in the dispute is as follows:

[0036] Time sequence extraction of diagnosis and treatment operation nodes, examination result nodes, and communication confirmation nodes in the dispute event chain;

[0037] Each diagnostic and treatment node is matched with the standard diagnostic and treatment path in the clinical diagnostic and treatment knowledge base to identify behavioral segments that differ from each other;

[0038] Extract the corresponding time offset state, operation sequence change state, and information missing state for the differential behavior segments;

[0039] By constructing a diagnostic and treatment comparison analysis structure based on the degree of behavioral deviation and the degree of norm compliance, corresponding objective diagnostic and treatment analysis results are formed.

[0040] By utilizing a virtual third-party evaluator role, the analysis results are subjected to neutral semantic transformation, and objective explanations that conform to the mediation scenario are output.

[0041] Preferably, the process of simultaneously generating multiple versions of on-site modulated scripts is as follows:

[0042] Semantic segmentation is performed on the requests submitted by both doctors and patients to extract frequently disputed statements and emotionally sensitive expressions;

[0043] Based on the semantic relationships among frequently disputed statements, clustering and labeling of cognitive biases in diagnosis and treatment in disputes;

[0044] By combining the density of technical terms and the intensity of emotional conflict in the clustering results, the medical explanations are stratified and transformed.

[0045] Generate empathic, professional, and popular versions of mediation scripts based on different mediation scenarios, and differentiate the tone intensity and explanation depth in the scripts.

[0046] By simulating communication feedback results, different versions of the script are adapted and adjusted to form the output of on-site mediation script.

[0047] Preferably, the process for generating a standardized review report with a compliance disclaimer is as follows:

[0048] The facts of diagnosis and treatment, the points of contention, and the results of mediation and communication in the chain of disputes are summarized in a structured manner.

[0049] Based on the results of comparison with the treatment guidelines and the results of the dispute analysis, corresponding neutral review content is generated;

[0050] Filter out biased descriptions, statements implying liability, and content that infers compensation from the evaluation content;

[0051] Embed a compliance disclaimer in the document to explain the system's role, the scope of application of the opinion, and its non-judicial nature, and output a standardized evaluation opinion in accordance with a unified document template.

[0052] Preferably, the process of archiving dispute resolution procedures without privacy protection is as follows:

[0053] De-identification processing is performed on dispute data that has been mediated, removing patient identity information and medical institution-related information;

[0054] Anonymized case tags are constructed based on the type of dispute, the focus of the controversy, and the differences in medical treatment. Communication records, wording adjustments, and evaluation results during the mediation process are classified and archived.

[0055] High-frequency dispute patterns are identified through case similarity analysis, and corresponding dispute risk association structures are formed. Anonymized dispute analysis data is then generated using these dispute risk association structures.

[0056] Compared with existing technologies, the present invention provides an AI-based virtual third-party neutral review and mediation method and system for medical disputes, which has the following beneficial effects:

[0057] By constructing an AI-powered virtual third-party neutral evaluation and mediation system independent of medical institutions and patients, a unified correlation analysis of treatment behaviors, communication behaviors, and disputed behaviors in the process of handling medical disputes is achieved. Compared with traditional manual mediation methods, it can objectively and neutrally analyze the focus of disputes in complex medical disputes without directly intervening in the determination of medical liability and compensation. The data anonymization module performs structured anonymization and consistency verification on private information in medical records, examination reports, and communication records, effectively reducing the risk of leakage of sensitive patient information and improving privacy security during the data flow process. The treatment standard comparison module compares the actual treatment process with the standard treatment path item by item, quickly identifying behavioral differences and missing information points in the treatment process, improving the dispute analysis process. The system improves efficiency in identifying problems during the process; it simplifies professional medical content through a dispute analysis module, enabling patients to more intuitively understand diagnosis and treatment logic, reducing communication barriers caused by differences in professional knowledge; simultaneously, the mediation script generation module dynamically adjusts the order of explanation, intensity of expression, and reassurance strategies based on the emotional state of both doctors and patients and communication feedback, improving communication adaptability and emotional mitigation during mediation; the evaluation document output module automatically generates standardized neutral evaluation opinions with compliance disclaimers, and supports various commercial adaptation models such as single-case services, annual packaged services, and regional centralized procurement services, enabling rapid access without additional hardware deployment, thereby improving the system's deployment flexibility, service coverage, and practical application efficiency in different medical institutions and regional dispute mediation scenarios. Attached Figure Description

[0058] Figure 1 This is a schematic diagram of the structure of the present invention;

[0059] Figure 2 This is a schematic diagram of the method of the present invention. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] Example 1: Please refer to Figure 1 The AI ​​virtual third-party neutral review and mediation system for medical disputes in this embodiment of the invention includes:

[0062] Data anonymization module: Receives medical dispute materials entered by the hospital, performs structured parsing of medical record texts, examination reports, treatment records, doctor-patient communication content, and auxiliary examination data, automatically identifies and masks patient identity information, contact information, and related privacy fields, and completes data anonymization verification.

[0063] The data anonymization and verification process in the data anonymization module is as follows:

[0064] The system performs structured parsing on medical records, examination reports, treatment records, doctor-patient communication information, and supporting materials entered by the hospital, extracting corresponding identity fields, time-related fields, and treatment-related fields. It also receives electronic medical records, laboratory reports, imaging examination descriptions, surgical records, nursing records, and transcribed doctor-patient communication texts uploaded by the hospital, and performs unified parsing processing on data of different formats. For text materials, it uses field segmentation to extract patient name, hospital number, department information, treatment time, and examination items. For tabular or semi-structured materials, it performs position matching of time items, examination items, and treatment items according to preset field templates. For continuous communication records, it establishes a corresponding communication sequence according to the speaking time order. After parsing, the patient identity-related fields, treatment process-related fields, and time-related fields are stored in the corresponding data mapping areas.

[0065] Based on entity recognition results, tiered masking is applied to patient names, ID card numbers, contact numbers, home addresses, and hospitalization numbers. Using pre-defined entity recognition rules, name strings, numeric strings, and address descriptions in structured data are scanned. When an ID card number is recognized, the first few area codes are retained while the remaining digits are hidden. When a contact number is recognized, only part of the area code information is retained. For the name field, an anonymous replacement identifier is used. For hospitalization numbers, bed numbers, and examination serial numbers, different levels of hiding are applied based on the importance of each field. Identity information present in image descriptions, attachment notes, and free descriptions is also masked to prevent patient identity leakage through non-standard fields.

[0066] Cross-mapping relationships are established for recurring related identity information in different medical materials, and uniform replacement processing is performed on the mapped identity fields. Identity fields in medical records, examination request forms, surgical records, and communication materials are repeatedly compared, and a unified identity mapping index is established based on the relationship between name, time point, and treatment items. When the same patient has abbreviations, aliases, or partially hidden identity information in different materials, they are merged into the same anonymous identity label through the association index. The unified replacement rules are used to synchronously replace the identity fields in all related materials with consistent anonymous identifier content, so that the case relevance can be maintained between different materials, while avoiding identity re-identification due to differences in field expression.

[0067] Semantic association scanning identifies implicit descriptions that may reveal a patient's identity, and these descriptions are then reconstructed by blurring. Semantic association analysis is performed on natural language content in doctor-patient communication records, free descriptions of illness, and special treatment notes to identify information fragments that may indirectly reveal a patient's identity. For example, descriptions containing specific identity markers such as "a teacher at a certain school," "a person in charge of a certain company," or "the only caregiver for an elderly person" are risk-marked. For descriptions containing specific times, specific locations, or combinations of rare diseases, contextual semantic analysis is used to assess the risk of identity recognition. When the identification results exceed a preset risk threshold, a generalized replacement method is used to blur the relevant descriptions, replacing specific occupations with general occupational categories, precise locations with regional expressions, and performing interval processing on specific time information to reduce the possibility of the patient being indirectly identified.

[0068] By combining the proportion of residual fields, semantic integrity, and information association consistency in the anonymized data, a verification analysis is performed on the anonymization results to form data anonymization results that meet privacy protection requirements. After the anonymization process is completed, the number of unmasked fields in the anonymized text is counted, and the proportion of residual identity information is calculated. Semantic integrity analysis is performed on the anonymized diagnosis and treatment content to detect whether the deletion of fields has caused a break in the diagnosis and treatment logic, missing event chains, or unrelated contexts. For anonymous fields between different materials, their mapping consistency is further verified to avoid situations where the same patient corresponds to multiple anonymous labels or different patients share the same anonymous label. When residual risk fields, semantic missing, or mapping anomalies are detected, the anonymization correction of the corresponding fields is automatically re-executed. After verification, data anonymization results that meet privacy protection requirements and maintain the correlation of diagnosis and treatment events are output.

[0069] The diagnostic and treatment standard comparison module connects to a general clinical diagnostic and treatment knowledge base, objectively comparing the patient reception behavior, examination behavior, treatment behavior and disclosure behavior involved in the dispute with the industry's routine diagnostic and treatment standards, and identifying behavioral differences and missing information nodes in the diagnostic and treatment process.

[0070] The process of identifying behavioral discrepancies and missing information nodes in the diagnosis and treatment process in the diagnosis and treatment guideline comparison module is as follows:

[0071] The time sequence of consultation records, examination records, treatment records, and communication records involved in the dispute is organized to form a corresponding chain of diagnosis and treatment behaviors. The outpatient records, inpatient medical records, medical order execution records, examination request forms, nursing records, and doctor-patient communication logs uploaded by the hospital are uniformly accessed and sorted according to the timestamp information in each record. For records with different data sources, corresponding event tags are established according to the behavior categories such as consultation, examination, treatment, follow-up examination, and notification. When there are multiple parallel diagnosis and treatment behaviors in the same time period, a sequential correlation structure is established based on the behavior dependency relationship. For records without clear time marks, the corresponding time interval is inferred by combining the context of diagnosis and treatment content and related examination results. After completing the time organization, a chain of diagnosis and treatment behaviors covering the patient's complete diagnosis and treatment process is formed.

[0072] Each node in the treatment behavior chain is matched item by item with the standard treatment path in the clinical treatment knowledge base to identify treatment behaviors with sequence deviations, missing content, or abnormal execution intervals. The pre-defined clinical treatment knowledge base is invoked to extract the standard treatment path corresponding to the current disease type, treatment method, and examination items. The reception node, examination node, treatment node, and postoperative observation node in the treatment behavior chain are compared item by item with the corresponding nodes in the standard path. When the actual treatment behavior involves examinations performed earlier or later than scheduled, key examinations not performed, treatment steps missing, or the execution time of medical orders exceeding the standard interval, the corresponding behavior is marked as abnormal. For collaborative treatment scenarios between different departments, the analysis further examines whether there are abnormal delays in the treatment connection time between different departments. Through the item-by-item matching process, treatment behavior deviation results are generated.

[0073] Based on the time span, operational continuity, and treatment dependency between behavioral nodes, abnormal behavioral segments are clustered to form corresponding behavioral difference nodes. Identified abnormal behaviors are analyzed for correlation, calculating the time interval, operational continuity, and treatment logic correlation between different abnormal behaviors. When multiple abnormal behaviors occur consecutively within the same treatment stage, they are grouped into the same abnormal behavioral segment. For situations such as treatment being implemented directly after missed examinations, operations being performed without complete notification, or treatment plans being adjusted without examination results being returned, behavioral association chains are established based on treatment dependency. Behavioral segments with the same abnormal type or the same impact path are grouped through clustering to form corresponding behavioral difference nodes. Each behavioral difference node reflects key deviation areas in the treatment process that may lead to doctor-patient disputes.

[0074] A completeness scan is performed on the informed consent content, signing records, and examination feedback results in the treatment process chain to identify unrecorded, unfeeded, or interrupted process areas, forming corresponding missing information nodes. Completeness checks are also performed on informed consent forms, risk disclosure records, examination result feedback records, and communication confirmation records in the treatment process chain. When there are unsigned informed consent documents, unfeeded examination results, or missing clear communication records, the corresponding areas are marked as missing information areas. For process segments with abnormal time spans and no supplementary records, they are further identified as interrupted information areas. Combined with the necessary informed consent requirements in the standard treatment pathway, the completeness of informed consent in the actual treatment process is compared and analyzed. After completing the scan, unrecorded nodes, unfeeded nodes, and interrupted information nodes are collectively summarized as missing information nodes, and corresponding missing information analysis results are generated.

[0075] The Dispute Analysis Module extracts the core demands and disagreements between doctors and patients, conducts correlation analysis on frequently disputed statements, emotionally conflicting statements, and content related to cognitive biases in diagnosis and treatment in dispute materials, and transforms professional medical content into popular interpretations.

[0076] The process of transforming professional medical content into popular interpretations in the controversy analysis module is as follows:

[0077] The medical terminology, descriptions of diagnosis and treatment, and explanations of examination results output by the dispute resolution module are semantically segmented to extract professional expression fragments. Natural language parsing is performed on the generated text of diagnosis and treatment disputes to identify professional terms related to disease names, examination indicators, treatment plans, drug instructions, and risk disclosures. For continuous descriptions of diagnosis and treatment, the text is segmented according to the semantic structure of "diagnosis and treatment—reason explanation—result feedback." When abbreviations, professional codes, or industry-standard expressions are present in the text, a pre-defined medical terminology dictionary is used for corresponding expansion. For indicator data and professional conclusions in examination reports, the corresponding indicator names, reference ranges, and abnormality explanations are extracted. After segmentation, each professional expression fragment is categorized and stored according to diagnosis, treatment, examination, and risk categories.

[0078] Based on pre-defined medical terminology mapping rules, the clinical meaning, treatment purpose, and risk descriptions corresponding to professional expression fragments are semantically replaced with more accessible language. A terminology mapping database containing the correspondence between medical terms, clinical explanations, and commonly used public expressions is established. When a professional term is identified, it is replaced with a description that is easy for patients to understand. For example, "local tissue ischemia" is replaced with "local blood supply insufficiency," and "conservative treatment" is replaced with "treatment methods that temporarily refrain from surgery." For abnormal indicators in examination results, corresponding descriptions of physical effects and possible symptoms are added simultaneously. For descriptions of treatment purposes, everyday expressions are used to explain the actual effects of the treatment. For professional descriptions involving risks, possible consequences are explained in a probabilistic and scenario-based manner, avoiding the use of absolute medical conclusions. Through the semantic replacement process, professional medical content can be transformed into explanations that are suitable for ordinary patients to understand.

[0079] The system dynamically adjusts the level of expression and complexity of simplified explanations based on the emotional fluctuations and cognitive differences between doctors and patients during communication. It analyzes the speech-to-text, repeated questions, and emotional expressions used by both parties to identify their current level of emotional tension and comprehension. When patients exhibit significant anxiety, skepticism, or resistance, the system reduces the density of technical terms in the explanations and prioritizes concise sentences and reassuring expressions. When patients possess higher medical cognitive abilities, some technical concepts are appropriately retained, and explanations of treatment principles are added. For points of contention where there are significant cognitive discrepancies between doctors and patients, the system unfolds the treatment logic step-by-step, dynamically adjusting sentence length, explanation order, and expression intensity based on real-time communication feedback to improve comprehension efficiency and acceptance during communication.

[0080] To address the potential for misunderstandings in medical practice descriptions, a scenario-based explanatory structure is introduced to link the reasons for treatment, the process, and the outcome. For controversial medical practices such as "failure to conduct timely examinations," "adjustment of treatment plans," and "failure to immediately perform surgery," corresponding scenario-based explanatory templates are established. During the explanation process, the patient's condition at the time of the medical practice, the basis of the clinical judgment, the available treatment options at the time, and the subsequent treatment results are displayed in a linked manner. For example, for delayed examinations, the system explains the reasons for the delay, the purpose of the observation, and the patient's vital signs at the time; for adjustments to medication plans, it explains the changes in the original treatment effect and the expected effect after the adjustment. A timeline is used to display the continuous changes before and after the medical practice, enabling patients to understand the relationship between medical actions and their clinical state, thereby reducing the risk of misunderstandings caused by single behavioral fragments.

[0081] The system performs logical corrections on the converted explanations through semantic consistency checks, resulting in simplified interpretations suitable for medical dispute mediation scenarios. Semantic consistency checks are then performed on the simplified explanation text to analyze for logical conflicts, inconsistencies, or ambiguities between different explanation paragraphs. For statements containing absolute judgments, implied responsibility, or potential for misinterpretation, the system automatically performs neutralization corrections. Cross-validation is conducted between the simplified explanations and the original medical records to prevent deviations in medical meaning during simplification. For explanations involving complex medical procedures, further checks are performed to identify omissions of key steps or breaks in causal relationships. After logical corrections, simplified interpretations suitable for medical dispute mediation scenarios are output, ensuring that the explanations are easy to understand while maintaining consistency and neutrality in the medical facts.

[0082] Mediation script generation module: Based on the type of dispute, the intensity of the controversy, and the emotional state of the doctor and patient, it automatically generates three types of on-site mediation scripts: empathic version, professional version, and popular version. It also dynamically adjusts the order of explanation, intensity of expression, and reassurance strategies based on communication feedback.

[0083] The process of dynamically adjusting the order of explanations, intensity of expression, and reassurance strategies based on communication feedback in the mediation script generation module is as follows:

[0084] The voice and text recordings, emotional feedback, and interactive response results of both doctors and patients during the mediation process are continuously collected, and the corresponding emotional fluctuation features and communication sensitivity features are extracted.

[0085] Based on the duration of time both the doctor and the patient spends on different explanations, the frequency of repeated questioning, and the degree of semantic conflict, identify the highly sensitive points of contention in the current mediation stage;

[0086] Based on the relevant position of highly sensitive points of contention in the chain of disputes, the order of explanations in mediation scripts is dynamically rearranged, prioritizing explanations that are more relevant to emotional relief.

[0087] Based on the range of emotional changes and the degree of acceptance of communication between the doctor and the patient, the tone intensity, density of professional terminology, and proportion of reassuring expressions in the mediation script are adaptively adjusted.

[0088] The evaluation document output module integrates the results of comparison with treatment guidelines, analysis of points of contention, and mediation and communication results in a structured manner to generate a standardized neutral evaluation opinion. The document includes a built-in compliance disclaimer and supports three commercial adaptation modes: single case service, annual package service, and regional centralized procurement service.

[0089] The process of integrating a compliance disclaimer into the evaluation document output module and supporting three commercial adaptation modes—single case service, annual package service, and regional centralized procurement service—is as follows:

[0090] The results of comparison with treatment guidelines, analysis of points of contention, and mediation and communication records are structured and integrated to form a corresponding set of review documents;

[0091] By combining the compliance judgment boundaries in the neutral role preset module, the responsibility-oriented statements, compensation-implying statements and absolute judgment content in the review documents are filtered out, and a standardized compliance disclaimer is embedded in the documents;

[0092] Based on the case handling scale, service cycle requirements, and regional collaboration scope of medical institutions, the system service mode is configured as a single case service mode, an annual package service mode, or a regional centralized procurement service mode.

[0093] Based on the frequency of case calls, the number of documents generated, and the number of institutions accessing the service in different commercialization adaptation models, the output permissions of review documents, the scope of data access, and the service resource allocation strategy are dynamically adjusted to form corresponding commercialization service adaptation results.

[0094] Example 2: As Figure 2 As shown, the AI-powered virtual third-party neutral review and mediation method for medical disputes includes the following steps:

[0095] S1: Accept information on medical disputes, treatment processes, and the demands of both parties entered by medical institutions, initiate privacy data desensitization verification, and remove all personal identity information.

[0096] The process of removing all personally identifiable information in S1 is as follows:

[0097] The system performs structured parsing on medical records, examination reports, treatment records, and doctor-patient communication materials entered by medical institutions, extracting patient identity fields, treatment association fields, and time stamp fields.

[0098] Based on the entity recognition results, tiered masking is applied to patient names, ID numbers, contact information, home addresses, and hospitalization numbers.

[0099] Establish cross-mapping relationships for identity-related content that appears repeatedly in different medical records, and perform unified replacement and anonymization processing on the mapped identity information;

[0100] Semantic association analysis was used to identify implicit descriptive fragments that might indirectly reveal the patient's identity, and these implicit descriptive fragments were then reconstructed by blurring.

[0101] By combining the field retention rate, semantic integrity, and data association consistency in the anonymized data, a verification analysis is performed on the anonymized results to generate dispute data without personal identification information.

[0102] S2: Invoke the preset virtual third-party neutral review role, and conduct an objective comparative analysis of the medical behaviors involved in the dispute in conjunction with clinical diagnosis and treatment guidelines.

[0103] The process of conducting an objective comparative analysis of the medical treatment involved in the dispute in S2 is as follows:

[0104] Time sequence extraction of diagnosis and treatment operation nodes, examination result nodes, and communication confirmation nodes in the dispute event chain;

[0105] Each diagnostic and treatment node is matched with the standard diagnostic and treatment path in the clinical diagnostic and treatment knowledge base to identify behavioral segments that differ from each other;

[0106] Extract the corresponding time offset state, operation sequence change state, and information missing state for the differential behavior segments;

[0107] By constructing a diagnostic and treatment comparison analysis structure based on the degree of behavioral deviation and the degree of norm compliance, corresponding objective diagnostic and treatment analysis results are formed.

[0108] By utilizing a virtual third-party evaluator role, the analysis results are subjected to neutral semantic transformation, and objective explanations that conform to the mediation scenario are output.

[0109] S3: Extract the core points of contention between doctors and patients, transform professional medical knowledge into popular interpretations, and simultaneously generate multiple versions of on-site scripts.

[0110] The process of simultaneously generating multiple versions of on-site modulated scripts in S3 is as follows:

[0111] Semantic segmentation is performed on the requests submitted by both doctors and patients to extract frequently disputed statements and emotionally sensitive expressions;

[0112] Based on the semantic relationships among frequently disputed statements, clustering and labeling of cognitive biases in diagnosis and treatment in disputes;

[0113] By combining the density of technical terms and the intensity of emotional conflict in the clustering results, the medical explanations are stratified and transformed.

[0114] Generate empathic, professional, and popular versions of mediation scripts based on different mediation scenarios, and differentiate the tone intensity and explanation depth in the scripts.

[0115] By simulating communication feedback results, different versions of the script are adapted and adjusted to form the output of on-site mediation script.

[0116] S4: Generate a standardized review report with a compliance disclaimer for use by medical institutions in on-site mediation.

[0117] The process of generating a standardized review report with a compliance disclaimer in S4 is as follows:

[0118] The facts of diagnosis and treatment, the points of contention, and the results of mediation and communication in the chain of disputes are summarized in a structured manner.

[0119] Based on the results of comparison with the treatment guidelines and the results of the dispute analysis, corresponding neutral review content is generated;

[0120] Filter out biased descriptions, statements implying liability, and content that infers compensation from the evaluation content;

[0121] Embed a compliance disclaimer in the document to explain the system's role, the scope of application of the opinion, and its non-judicial nature, and output a standardized evaluation opinion in accordance with a unified document template.

[0122] S5: Archive the dispute resolution process without privacy settings to complete the closed-loop dispute resolution process.

[0123] The process of archiving dispute resolution procedures without privacy in S5 is as follows:

[0124] De-identification processing is performed on dispute data that has been mediated, removing patient identity information and medical institution-related information;

[0125] Anonymized case tags are constructed based on the type of dispute, the focus of the controversy, and the differences in medical treatment. Communication records, wording adjustments, and evaluation results during the mediation process are classified and archived.

[0126] High-frequency dispute patterns are identified through case similarity analysis, and corresponding dispute risk association structures are formed. Anonymized dispute analysis data is then generated using these dispute risk association structures.

[0127] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0128] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An AI-powered virtual third-party neutral review and mediation system for medical disputes, characterized in that: include: Neutral Role Preset Module: Sets up virtual third-party evaluator roles that are independent of medical institutions and have no conflict of interest, and constructs corresponding neutral expression rules, emotional mitigation rules and compliance judgment boundaries. It imposes bias restrictions on the output content during the mediation process and does not make judgments on medical malpractice, liability ratios and compensation amounts. Data anonymization module: Receives medical dispute materials entered by the hospital, performs structured parsing of medical record texts, examination reports, treatment records, doctor-patient communication content and auxiliary examination data, automatically identifies and masks patient identity information, contact information and related privacy fields, and completes data anonymization verification; The diagnostic and treatment standard comparison module connects to a general clinical diagnostic and treatment knowledge base, objectively comparing the patient reception behavior, examination behavior, treatment behavior and disclosure behavior involved in the dispute with the industry's routine diagnostic and treatment standards, and identifying behavioral differences and missing information nodes in the diagnostic and treatment process; The Dispute Analysis Module extracts the core demands and conflicts between doctors and patients, conducts correlation analysis on frequently disputed statements, emotionally conflicting statements, and content related to cognitive biases in diagnosis and treatment in dispute materials, and transforms professional medical content into popular interpretations. Mediation script generation module: Based on the type of dispute, the intensity of the controversy, and the emotional state of the doctor and patient, it automatically generates three types of on-site mediation scripts: empathic version, professional version, and popular version. It also dynamically adjusts the order of explanation, intensity of expression, and reassurance strategies based on communication feedback. The evaluation document output module integrates the results of comparison with treatment guidelines, analysis of points of contention, and mediation and communication results in a structured manner to generate a standardized neutral evaluation opinion. The document includes a built-in compliance disclaimer and supports three commercial adaptation modes: single case service, annual package service, and regional centralized procurement service.

2. The AI ​​virtual third-party neutral evaluation and mediation system for medical disputes according to claim 1, characterized in that, The process of completing data anonymization and verification is as follows: The system performs structured parsing on medical records, examination reports, treatment records, doctor-patient communication information, and supporting materials entered by the hospital, and extracts the corresponding identity fields, time-related fields, and treatment-related fields. Based on the entity recognition results, tiered masking is applied to the patient's name, ID number, contact number, home address, and hospitalization number; Establish cross-mapping relationships for recurring related identity information in different medical records, and perform unified replacement processing on the mapped identity fields; Semantic association scanning is used to identify implicit descriptions that reveal patient identity, and these implicit descriptions are then reconstructed by blurring. By combining the proportion of fields remaining in the anonymized data, the degree of semantic integrity, and the consistency of information association, the anonymization results are validated and analyzed to form data anonymization results that meet privacy protection requirements.

3. The AI ​​virtual third-party neutral evaluation and mediation system for medical disputes according to claim 1, characterized in that, The process of identifying behavioral discrepancies and missing information points in the diagnosis and treatment process is as follows: Organize the patient reception records, examination records, treatment records, and communication records involved in the dispute into a time sequence to form a corresponding chain of medical behavior; Each behavioral node in the diagnosis and treatment behavior chain is matched with the standard diagnosis and treatment path in the clinical diagnosis and treatment knowledge base to identify diagnosis and treatment behaviors with sequence deviations, missing content, or abnormal execution intervals. Based on the time span, operational continuity, and diagnostic dependence among behavioral nodes, abnormal behavioral fragments are clustered to form corresponding behavioral difference nodes; Perform a completeness scan on the information provided, signed records, and examination feedback results in the diagnosis and treatment process chain to identify unrecorded, unfeasible, or interrupted process areas and form corresponding missing information nodes.

4. The AI ​​virtual third-party neutral review and mediation system for medical disputes according to claim 1, characterized in that, The process of transforming professional medical content into accessible explanations is as follows: Semantic segmentation is performed on the medical terminology, descriptions of diagnosis and treatment, and explanations of examination results output by the dispute resolution module to extract the professional expression fragments; Based on the preset medical terminology mapping rules, the clinical meaning, treatment purpose, and risk description of professional expression fragments are replaced with popular semantics. Based on the emotional fluctuations and cognitive differences between doctors and patients during the communication process, the expression level and sentence complexity of the simplified explanations are dynamically adjusted. To address the descriptions of medical procedures that are prone to misunderstanding, a scenario-based explanatory structure is introduced to provide a related explanation of the reasons for diagnosis and treatment, the process, and the results. The semantic consistency check is used to logically correct the converted interpretation, resulting in a simplified interpretation suitable for doctor-patient mediation scenarios.

5. An AI-powered virtual third-party neutral review and mediation method for medical disputes, applied to the AI-powered virtual third-party neutral review and mediation system for medical disputes as described in any one of claims 1-4, characterized in that, Includes the following steps: Upon receiving information from medical institutions regarding medical disputes, treatment processes, and the demands of both parties, a privacy data anonymization verification process is initiated to remove all personally identifiable information. By invoking a pre-defined virtual third-party neutral review role and combining it with clinical treatment guidelines, an objective comparative analysis of the treatment behaviors involved in the dispute is conducted. Extract the core points of contention between doctors and patients, transform professional medical knowledge into popular interpretations, and simultaneously generate multiple versions of on-site scripts for adjustment. Generate standardized review reports with compliance disclaimers for use by medical institutions in on-site mediation; The dispute resolution process is archived without privacy restrictions to complete the closed-loop dispute resolution process.

6. The AI ​​virtual third-party neutral review and mediation method for medical disputes according to claim 5, characterized in that, The process of conducting an objective comparative analysis of the medical treatment involved in the dispute is as follows: Time sequence extraction of diagnosis and treatment operation nodes, examination result nodes, and communication confirmation nodes in the dispute event chain; Each diagnostic and treatment node is matched with the standard diagnostic and treatment path in the clinical diagnostic and treatment knowledge base to identify behavioral segments that differ from each other; Extract the corresponding time offset state, operation sequence change state, and information missing state for the differential behavior segments; By constructing a diagnostic and treatment comparison analysis structure based on the degree of behavioral deviation and the degree of norm compliance, corresponding objective diagnostic and treatment analysis results are formed. By utilizing a virtual third-party evaluator role, the analysis results are subjected to neutral semantic transformation, and objective explanations that conform to the mediation scenario are output.

7. The AI ​​virtual third-party neutral evaluation and mediation method for medical disputes according to claim 5, characterized in that, The process of simultaneously generating multiple versions of on-site modulated scripts is as follows: Semantic segmentation is performed on the requests submitted by both doctors and patients to extract frequently disputed statements and emotionally sensitive expressions; Based on the semantic relationships among frequently disputed statements, clustering and labeling of cognitive biases in diagnosis and treatment in disputes; By combining the density of technical terms and the intensity of emotional conflict in the clustering results, the medical explanations are stratified and transformed. Generate empathic, professional, and popular versions of mediation scripts based on different mediation scenarios, and differentiate the tone intensity and explanation depth in the scripts. By simulating communication feedback results, different versions of the script are adapted and adjusted to form the output of on-site mediation script.

8. The AI ​​virtual third-party neutral review and mediation method for medical disputes according to claim 5, characterized in that, The process of generating a standardized review report with a compliance disclaimer is as follows: The facts of diagnosis and treatment, the points of contention, and the results of mediation and communication in the chain of disputes are summarized in a structured manner. Based on the results of comparison with the treatment guidelines and the results of the dispute analysis, corresponding neutral review content is generated; Filter out biased descriptions, statements implying liability, and content that infers compensation from the evaluation content; Embed a compliance disclaimer in the document to explain the system's role, the scope of application of the opinion, and its non-judicial nature, and output a standardized evaluation opinion in accordance with a unified document template.

9. The AI ​​virtual third-party neutral review and mediation method for medical disputes according to claim 5, characterized in that, The process of archiving dispute resolution procedures without privacy protection is as follows: De-identification processing is performed on dispute data that has been mediated, removing patient identity information and medical institution-related information; Anonymized case tags are constructed based on the type of dispute, the focus of the controversy, and the differences in medical treatment. Communication records, wording adjustments, and evaluation results during the mediation process are classified and archived. High-frequency dispute patterns are identified through case similarity analysis, and corresponding dispute risk association structures are formed. Anonymized dispute analysis data is then generated using these dispute risk association structures.

Images