Intelligent service method and system for chronic and special diseases

Abstract

The invention discloses an intelligent service method and system for chronic and special diseases, and relates to the technical field of medical care. National and local policy texts are received and are converted into structured rules through natural language processing, and a reimbursement association model is established; receiving patient data, marking a source, checking integrity, sending the complete data to an artificial intelligence module for auditing, and generating a reimbursement qualification result; a final result is generated through manual auditing, the state is updated, a source is returned, and a notification is generated when the result changes; according to the method, the policy text is converted into the structured supervision rule through natural language processing, and the reimbursement association model is established, so that the problem of low efficiency of traditional policy storage and manual matching is solved; the reimbursement proportions of different regions can be automatically matched, the applicable error rate of the cross-region medical service seeing proportion is reduced, and policy adjustment is synchronized within 24 hours; special patients are identified through the artificial intelligence module, the auditing period is shortened, and self-payment loss is reduced; and the material integrity is automatically verified, AI and manual cooperative auditing are carried out, and the overall auditing efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of health care, in particular to a special disease intelligent service method and system. BACKGROUND

[0002] The ideal special disease management needs to realize the closed loop of policy accurate landing-patient accurate identification-efficient reimbursement execution, which not only adapts to the differentiated policies of the state and the local, but also provides hierarchical protection for special patients. The high-burden patients whose self-payment amount accounts for more than 30% of the income need to submit materials such as income proof, and the manual audit cycle takes 3-5 days, and about 40% of the patients miss the priority reimbursement rights and interests due to untimely reporting; In addition, the patients need to manually submit the diagnosis materials of each disease, and the omission rate is about 25%, such as diabetic patients with coronary heart disease, if the coronary heart disease is not reported, only the single disease can be reimbursed, resulting in that the patients bear 10%-15% of the self-payment; The national and local policies are stored in the form of text, and the reimbursement proportion of the same disease in different regions (such as 80% of diabetes in city A and 75% in city B) needs to be manually queried, and the error rate of the proportion of patients who go to hospital in different regions is about 20%; After the local policy is adjusted, the reimbursement rules need to be manually updated for 7-14 days, and during this period, about 15% of the patients cannot enjoy the reimbursement preferential treatment brought by the new policy due to the unsynchronized rules. The existing system lacks the automatic analysis capability of unstructured data, and a dynamic association model between the policy and the patient is not established, resulting in that the special patients are not identified in time and the policy execution is biased. SUMMARY

[0003] The purpose of the present application is to provide a special disease intelligent service method and system to solve the problems in the background art.

[0004] In order to achieve the above purpose, the present application provides the following technical scheme: a special disease intelligent service method, comprising: Receiving the policy text input manually, the policy text including national rigid policy medical information and / or local policy medical information; Converting the policy text into structured regulatory rules by natural language processing technology, and establishing a reimbursement association model; Receiving patient data and marking the data source, and analyzing the integrity of the patient data; if the patient data is complete, the complete patient data is transmitted to an artificial intelligence module for auditing, and if the patient data is incomplete, the incomplete patient data is transmitted to the data source and the missing data information is prompted; The artificial intelligence module calls the reimbursement association model and the regulatory rules, audits the complete patient data, analyzes the patient disease information, and judges the patient's reimbursement qualification based on the disease information, and generates an audit reimbursement qualification result; The reimbursement eligibility results and the patient data are transmitted to the manual review port. After manual review and confirmation, the final reimbursement result is generated, and the patient's reimbursement eligibility status is updated according to the final reimbursement result. The final reimbursement result is transmitted back to the data source; and a notification message is generated when the final reimbursement result differs from the original reimbursement eligibility.

[0005] In a preferred embodiment, after receiving manually input policy text, a policy rule base is established, storing the policy text at the national policy-regional policy level: the national policy level specifically stores policies uniformly implemented nationwide, such as the "National Basic Medical Insurance Outpatient Chronic and Special Disease Catalog"; the regional policy level is categorized by province and city, storing "Outpatient Chronic and Special Disease Reimbursement Policy of Province B" and "Supplementary Outpatient Chronic and Special Disease Reimbursement Policy of City B, Province B" respectively, facilitating quick searching by level; the policy text is parsed using natural language processing technology, extracting "hypertension" from "Province B stipulates that the outpatient reimbursement ratio for hypertension patients is 70%" The key information of "-70%- Province B" – disease name, reimbursement ratio, and applicable region – is transformed into structured regulatory rules containing fields such as disease code, region code, and reimbursement ratio, and stored in the database. Based on the extracted key information, a reimbursement association model of region-disease-reimbursement ratio is established. When a manually input text indicating a local policy adjustment, such as "the reimbursement ratio for hypertension in Province B is adjusted to 75%", is received, natural language processing technology is used to automatically locate the disease, region, and new ratio involved in the adjustment, and automatically update the reimbursement ratio association relationship for hypertension in Province B in the model, ensuring that the model data is synchronized with the current policy.

[0006] In a preferred embodiment, when the artificial intelligence module reviews complete patient data, it parses the electronic medical records in the complete patient data using natural language processing technology, extracting keywords such as "hypertension" and "well controlled" from the description "the patient has sought medical treatment for hypertension in the past six months, and the blood pressure is well controlled." The extracted keywords are then compared and analyzed with the patient's hypertension treatment records over the past two years to determine the patient's disease status as "existing." The reimbursement association model and the regulatory rules are invoked, and based on the "existing" disease status and the patient's province (B), a corresponding reimbursement ratio of 70% is matched. Based on the 70% reimbursement ratio and the "existing" disease status, combined with the patient's medical treatment at a designated hospital in province B, the system determines the patient's eligibility for reimbursement, generating a reimbursement eligibility review result that includes the disease status, reimbursement ratio, eligibility determination result, and the basis for determination.

[0007] In a preferred embodiment, when the artificial intelligence module reviews complete patient data, it calculates the economic burden ratio based on the patient's out-of-pocket expenses for hypertension outpatient services in the past three months (3600 yuan) and the integrated monthly income data (4000 yuan). Since the ratio reaches 30%, the patient is marked as a high-economic-burden patient. Combining the family information in the electronic medical record ("the patient is from a single-parent family, raising one child alone") and relevant data provided by the community, the module extracts the family's special circumstances. According to preset scoring standards, it adds 3 points for new illness, 3 points for high economic burden, and 2 points for single-parent family, etc., to score the patient and generate a special patient level label. Based on the generated level label, it matches the patient with a corresponding priority reimbursement process, such as a secondary label corresponding to a 48-hour priority reimbursement process, accelerating the reimbursement process.

[0008] In a preferred embodiment, for patients marked as having a high economic burden, the artificial intelligence module filters out external assistance programs for the patient's hypertension from the system's stored assistance program library, such as "Medical Assistance for Chronic and Special Diseases in Province B," and automatically pushes assistance application guidelines that include the application conditions for the program (such as a family's monthly income being lower than the local minimum living standard), required materials (such as proof of income and a list of medical expenses), and application channels (such as a link to an online application platform). Existing information such as the patient's name, ID number, and hypertension diagnosis certificate is extracted from the patient's medical insurance reimbursement materials, and this information is reused to generate an assistance application template. Content in the template that requires the patient to supplement, such as "family income situation," is marked with a "needs to be supplemented" prompt. Through an information interface with the assistance organization, the module obtains real-time progress information such as "submitted," "under review," and "approved" for the patient's assistance application and records this progress information synchronously in the "Assistance Record" of the patient's data file, allowing the patient to easily check and understand the information at any time.

[0009] In a preferred embodiment, the analysis of the completeness of the patient data includes: pre-setting a material verification list based on the chronic and special disease type of the hypertensive patient, the list explicitly including materials such as "electronic medical records for the past 3 months (which must record blood pressure control status)," "blood pressure test reports for the past month," and "hypertension diagnosis certificates (issued by secondary or higher-level hospitals)"; identifying the received patient data using text recognition technology, extracting key information such as material name, issuance time, and issuing institution, and comparing the identification results with the pre-set material verification list item by item; if the patient does not provide a "blood pressure test report for the past month," an "material missing list" is automatically generated, indicating the name of the missing material, the supplementary requirements (such as "a blood pressure test report issued by a regular medical institution within the past month is required"), and the deadline for supplementation (such as "before August 10, 2024"); and according to the patient data source mark (such as "source: HIS system of hospital in city B, province B"), the list is pushed to the corresponding hospital HIS system review interface, and a text message notification is sent to relevant hospital staff.

[0010] In a preferred embodiment, the generation of the final reimbursement result after manual review and confirmation specifically includes: the manual review portal synchronously displays the AI-generated reimbursement eligibility result and corresponding patient data in a visual interface. The left side of the interface highlights the "Pass" judgment result and the basis for "Patient's disease status is active and meets the reimbursement conditions for hypertension in Province B" in card form. The right side of the interface provides access to view patient data such as electronic medical records, treatment records, and cost details; reviewers can click to view the complete content. Reviewers can also annotate their opinions online using text input boxes and commonly used opinion options such as "Agree to AI review result" and "Requires supplementary materials." After selecting an option, further explanation can be provided. For example, selecting "Agree to AI review result" allows reviewers to add specific details. Enter "Patient materials are complete and meet reimbursement requirements"; if approved, a final reimbursement result will be generated, including the patient's name, ID number, hypertension type information, 70% reimbursement ratio, reviewer's name, and review time; if rejected, the reviewer must select the rejection reason category such as "incomplete materials" or "disease status mismatch" on the interface and fill in a specific explanation (e.g., "No electronic medical records for the past 3 months provided, therefore rejected"), generating a final rejection result including the rejection reason and reviewer information; after the final reimbursement result is generated, it will be automatically synchronized and linked to the patient's data, stored in the "Reimbursement Record" of the patient's data, and established a corresponding relationship with the patient's ID number, which can be directly retrieved and viewed when querying the patient's data later.

[0011] This invention also provides a smart service system for chronic and special diseases, comprising: Policy and Rules Module: Configured to receive national and local policies input manually, transform them into structured regulatory rules through natural language processing technology, and establish and update the reimbursement association model; Data receiving and marking module: configured to receive patient data and mark the data source, and transmit patient data to the material verification module; The materials verification module is configured to analyze the integrity of patient data, transmit complete patient data to the AI ​​review module, and transmit incomplete patient data to the data source and indicate the missing data information. The AI ​​review module is configured to call the reimbursement association model and regulatory rules to review complete patient data, analyze patient disease information and determine reimbursement eligibility, and generate reimbursement eligibility review results. The manual review module is configured to receive the reimbursement eligibility review results and patient data, generate the final reimbursement result after manual review and confirmation, and update the patient's reimbursement eligibility status based on the final reimbursement result. Results Feedback Module: Configured to send the final reimbursement result back to the data source and generate a notification message when the final reimbursement result differs from the original reimbursement eligibility.

[0012] In a preferred embodiment, the policy rules module includes a policy input submodule, a rule parsing submodule, and an association model submodule, wherein: Policy Input Submodule: Configured to provide a manual input interface, supporting the uploading, modification, and storage of policy texts; The rule parsing submodule is configured to extract region-disease-reimbursement ratio information using natural language processing technology and convert it into computer-executable logic. The association model submodule is configured to build a region-disease-reimbursement ratio association model based on the extracted information and automatically update the model when policies are adjusted.

[0013] In a preferred embodiment, the policy rules module includes: The AI-powered review module includes a disease information analysis submodule, a reimbursement ratio calculation submodule, and a result output submodule. The disease information analysis submodule is configured to parse electronic medical records using natural language processing technology to determine the patient's disease status. The reimbursement eligibility determination submodule is configured to call the reimbursement association model and regulatory rules, and determine the patient's reimbursement eligibility based on the disease status and region and the matching reimbursement ratio. The results generation submodule is configured to generate the results of the reimbursement eligibility review and link them to the patient data. The special patient service module includes an identification and triage submodule, a reimbursement process adaptation submodule, and an assistance guidance submodule: The identification and classification submodule is configured to calculate the ratio of out-of-pocket expenses to average monthly income based on patient data, and generate special patient classification labels by combining special family circumstances. Reimbursement process adaptation submodule: configured to match the corresponding priority reimbursement process based on the level label; The assistance guidance submodule is configured to push assistance application guidance to patients with high economic burden, generate assistance application templates, and track progress.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention uses natural language processing technology to transform policy texts into structured regulatory rules and establish a reimbursement association model, solving the problems of low efficiency caused by storing policies in text form and manual matching in traditional models. For the same disease reimbursement ratio in different regions, the system can automatically match the corresponding ratio with the patient's region, reducing the error rate of ratio application for patients seeking medical treatment across regions. When local policies are adjusted, there is no need to manually update the rules; the system automatically synchronizes through the model, shortening the time lag for new policy implementation from 7-14 days to within 24 hours, ensuring that patients can enjoy policy benefits in a timely manner.

[0015] This invention uses an artificial intelligence module to automatically analyze patient data, enabling the identification of special patients without manual reporting: it automatically marks patients with high economic burden based on the formula of out-of-pocket amount / average monthly income, shortening the review cycle from 3-5 days to within 2 hours, allowing some patients who previously missed out on their rights due to failure to report in time to enjoy priority reimbursement; it automatically parses electronic medical records to identify patients with overlapping diseases, avoiding the problem of omissions in manually submitted materials. For example, patients with diabetes and coronary heart disease can have their reimbursement ratio automatically calculated according to the overlapping disease rules, reducing some of their out-of-pocket expenses.

[0016] This invention features automatic verification of material integrity. Incomplete data is returned to the source platform in real time with missing information, avoiding repeated communication due to missing materials in traditional manual review. The collaboration between AI and human review improves basic review efficiency while ensuring accuracy through human confirmation. The final reimbursement result is simultaneously transmitted back to the data source and a notification is generated when eligibility changes, achieving a closed-loop process from data submission to review and feedback, thus improving overall review efficiency. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0018] Figure 1 This is a flowchart of the method of the present invention.

[0019] Figure 2 This is a system block diagram of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0021] Example 1, please refer to Figure 1 As shown in this embodiment, a smart service method for chronic and special diseases includes: S1. Receive manually input policy text, which includes national mandatory policy medical information and / or local policy medical information; S2. Transform the policy text into structured regulatory rules using natural language processing technology, and establish a reimbursement association model; S3. Receive patient data and mark the data source, and analyze the completeness of the patient data; if the patient data is complete, transmit the complete patient data to the artificial intelligence module for review; if the patient data is incomplete, transmit the incomplete patient data to the data source and indicate the missing data information. S4. The artificial intelligence module calls the reimbursement association model and the supervision rules to review the complete patient data, analyze the patient's disease information, and determine the patient's reimbursement eligibility based on the disease information, and generate the reimbursement eligibility review result. S5. Transmit the reimbursement eligibility verification results and the patient data to the manual review port. After manual review and confirmation, generate the final reimbursement result and update the patient's reimbursement eligibility status according to the final reimbursement result. S6. The final reimbursement result is transmitted back to the data source; and a notification message is generated when the final reimbursement result differs from the original reimbursement eligibility.

[0022] As described in steps S1-S6 above, with the rising incidence of chronic and special diseases such as hypertension and diabetes, and the increasing number of patients with overlapping illnesses and high economic burdens, higher requirements are placed on the accuracy and timeliness of medical insurance reimbursement. Ideal chronic disease management requires a closed loop of precise policy implementation, accurate patient identification, and efficient reimbursement execution. This necessitates adapting to differentiated national and local policies while providing tiered protection for special patients. High-burden patients with out-of-pocket expenses exceeding 30% of their income must submit income verification documents, with manual review taking 3-5 days. Approximately 40% of patients miss out on priority reimbursement due to delayed reporting. Patients with multiple conditions must manually submit diagnostic materials for each disease, resulting in an underreporting rate of approximately 25%. For example, a diabetic patient with coronary heart disease who fails to report coronary heart disease can only receive reimbursement for a single condition, leading to an additional 10%-15% out-of-pocket expense. National and local policies are stored in text form, requiring manual verification of reimbursement rates for the same condition in different regions (e.g., 80% for diabetes in City A, 75% in City B). The error rate for applying these rates to patients seeking treatment across regions is approximately 20%. After local policy adjustments, reimbursement rules need to be manually updated every 7-14 days. During this period, a small number of patients are unable to enjoy the reimbursement benefits of the new policy due to outdated rules. The existing system lacks the ability to automatically parse unstructured data and has not established a dynamic correlation model between policies and patients, resulting in untimely identification of special patients and deviations in policy implementation; By using natural language processing technology to transform policy texts into structured regulatory rules and establish reimbursement correlation models, the system can automatically match reimbursement ratios for the same disease in different regions, reducing the error rate in applying reimbursement ratios across regions. When local policies are adjusted, the system automatically synchronizes, reducing the implementation delay of new policies from 7-14 days to within 24 hours. Simultaneously, the artificial intelligence module automatically identifies special patients, reducing the review cycle for patients with high economic burdens from 3-5 days to 2 hours, and allowing a small number of patients who missed out on their rights to enjoy priority reimbursement. It also automatically identifies patients with overlapping diseases, reducing a small portion of out-of-pocket expenses. Furthermore, automatic verification of material completeness avoids repeated communication, and review results are fed back synchronously, achieving a closed-loop process and improving overall review efficiency. In one embodiment, future technology evolution is adapted; (1) Among them, the policy text acquisition and adaptation in step S1: This step supports both manual input and automatic retrieval modes: Current mode: Receives manually input policy texts, including national rigid policies (such as the "National Basic Medical Insurance Outpatient Chronic and Special Disease Catalog") and local differentiated policies (such as "C Province hypertension reimbursement rate 70%, D Province hypertension reimbursement rate 75%").

[0023] Future evolution model: Connect to the API interface of the policy release platform of the National Healthcare Security Administration and local healthcare security bureaus to automatically capture policy texts (such as obtaining the "National Unified Reimbursement Policy for Chronic and Special Diseases" through a structured data interface) without manual input; at the same time, retain the manual input channel for temporary policy adjustments or special policy supplements.

[0024] (2) Among them, in step S2, policy rule transformation and model adaptation: The acquired policy text is parsed using natural language processing technology. Current adaptation: Extract key information such as "region-disease-reimbursement ratio" (e.g., "Province C-diabetes-70%)", transform it into structured regulatory rules, and establish a reimbursement association model that includes the regional dimension.

[0025] Future Evolution and Adaptation: If a nationally unified policy text is received (such as "the reimbursement rate for diabetes is uniformly 75% nationwide"), the system will automatically identify the "no regional differences" characteristic and switch the reimbursement association model to the "nationally unified mode," adjusting the core dimension of the model to "disease type - reimbursement rate." If there is a "nationally unified + local transition" policy (such as "the nationally unified rate is 75%, while E province retains 80% during the transition period"), the model will automatically add a "transition identifier," prioritizing the matching of transition rates to patients in E province, and automatically switching to the nationally unified rate after the transition period ends.

[0026] (3) Among them, the future evolution adaptation scenarios in steps S3-S6: Taking "nationwide unified reimbursement policy + automatic policy retrieval" as an example: After the National Healthcare Security Administration released the "National Unified Reimbursement Policy for Chronic and Special Diseases", the system automatically captured the text through API, and natural language processing technology identified the "no regional differences" rule to generate the "disease type - unified ratio" regulatory rule. After patient data (such as medical records of diabetic patients) is pushed from the hospital's HIS system to the system, the material verification function automatically checks its completeness. The artificial intelligence module calls the "nationally unified model" and directly matches the "diabetes - 75%" reimbursement ratio without judging the region. It also determines the patient's eligibility for reimbursement by combining the "existence" status in the patient's electronic medical record. After manual review, the audit results will be simultaneously sent to the patient's official WeChat account. If the patient was originally reimbursed at a local rate of 65%, and this time is reimbursed at a nationally unified rate of 75%, the system will automatically generate a "Reimbursement Rate Adjustment Notice". As described in (1)-(3) above, the system uses a dual-mode approach of "manual input + automatic acquisition" to automatically capture policy texts by connecting to the policy release platform API of the medical insurance bureau, reducing the cost and error of manual input. Compared with the current pure manual input mode, the policy acquisition cycle is shortened from "1-2 working days" to "real-time synchronization". The manual input channel is retained to deal with special scenarios such as temporary policy adjustments and avoid blind spots in the system's adaptation to unconventional policies. At the policy rule conversion level, the system can automatically identify policy characteristics such as "nationally unified", "regional differences" and "transitional policies" and dynamically adjust the reimbursement association model dimensions (such as switching from "region-disease-ratio" to "disease-ratio"). For "nationally unified + local transition" policies, the model achieves differentiated adaptation through "transitional identifiers", which not only ensures the rigid implementation of nationally unified policies, but also accommodates special rules during the local transition period. After the transition period ends, it automatically switches to the unified ratio without manual intervention, reducing the policy switching error rate from "5%-8%" to "below 0.1%". Under the "nationally unified reimbursement policy" scenario, the system no longer needs to match and verify the patient's regional information. The AI ​​review process directly calls the "disease-unified ratio" model, reducing the number of review steps and shortening the review time per case from "5-10 minutes" to "2-3 minutes". The document verification and result feedback processes remain automated, and patients do not need to submit additional supporting documents due to regional differences, significantly improving the convenience of cross-regional medical reimbursement. When the reimbursement ratio changes due to policy adjustments (such as from 65% locally to a nationally unified 75%), the system automatically generates a "reimbursement ratio adjustment notification" and synchronizes it to the patient's terminal, clearly informing them of the reason for the adjustment, the old and new ratios, and the effective date. This solves the problem of patients being unaware of policy adjustments and reimbursement changes in the traditional model, greatly improving patients' awareness of reimbursement rules. In one embodiment, receiving the policy text and performing preliminary processing in step S1 specifically includes: S101. Receive manually input policy text, which includes national mandatory policy medical information (such as the "National Basic Medical Insurance Outpatient Chronic and Special Disease Catalog") and / or local policy medical information (such as "Province A Chronic and Special Disease Reimbursement Policy"). S102. Classify the received policy texts, distinguishing between national policies and regional policies, with regional policies further subdivided according to provincial and municipal administrative divisions; S103. Establish a policy rule database and store policy texts categorized by the "National Policy - Regional Policy" hierarchy. For example, store "Province A Chronic and Special Disease Reimbursement Policy" under the "Province Policy" directory and "Supplementary Regulations on Chronic and Special Disease Reimbursement in City A of Province A" under the "City Policy" directory. S104. Mark the entry time, entry personnel, and other metadata for each stored policy text, and standardize the text format. As described in steps S101-S104 above, by receiving manually input policy texts and clearly distinguishing between national and local policies, the system avoids query confusion caused by the mixed storage of policies at different levels. For example, the "National Basic Medical Insurance Outpatient Chronic and Special Disease Catalog" and "Province A Chronic and Special Disease Reimbursement Policy" are categorized separately, enabling quick location of the required policy level during subsequent retrieval, significantly improving policy search efficiency. The hierarchical storage structure of "National Policy - Regional Policy," combined with the detailed division at the provincial and municipal levels, ensures that the storage logic of policy texts is consistent with the administrative level. For instance, "Supplementary Regulations for Chronic and Special Disease Reimbursement in City A of Province A" can be directly linked to the "Province A" policy catalog, forming a complete policy framework. The marking of metadata such as entry time and entry personnel provides clear evidence for policy traceability. When there are questions about the policy content, the entry node and relevant responsible persons can be quickly located, reducing the traceability efficiency from the original 2 hours to within 10 minutes. After standardizing the text format, the font, paragraphs, and key information expression of policy texts remain consistent, avoiding natural language processing errors caused by format differences. For example, the term "reimbursement rate" is consistently expressed as "XX%" across all texts, significantly improving the accuracy of subsequent entity recognition models in extracting key information and reducing information omissions or misjudgments caused by inconsistent formatting. The hierarchical storage structure supports the dynamic addition or adjustment of policy texts based on administrative divisions. For instance, when adding "Chronic and Special Disease Policy for Province B," a corresponding storage node can be created directly under the provincial directory without reconstructing the overall storage framework. This scalability ensures the system can adapt to the differentiated management needs of policies in various regions, reserving room for future nationwide policy integration or local policy adjustments.

[0027] In one embodiment, the policy text transformation and model construction in step S2 specifically include: S201. Preprocess the policy text using natural language processing technology, including word segmentation, punctuation removal, and redundant modifier removal; S202. Use an entity recognition model to locate entities such as "diabetes", "70%" and "Province A", and use a relation extraction model to determine the relationship between the three (e.g., "diabetes-70%-Province A"). S203. Transform the extracted relationships into structured regulatory rules and store them in tabular form. The table contains fields such as "Disease ID", "Disease Name", "Regional Code" and "Reimbursement Ratio". S204. Establish a reimbursement association model based on structured regulatory rules, with the model using the region code and disease ID as the joint primary key to construct a hash table structure; S205. When receiving a text of local policy adjustment, the adjustment information is identified through natural language processing technology, the corresponding data in the model is automatically updated, and the new data is executed after the effective time. As described in steps S201-S205 above, preprocessing the policy text using natural language processing technology (word segmentation, punctuation removal, etc.) can filter out irrelevant information, allowing subsequent entity recognition to focus more on core content. For example, from the sentence "According to the latest regulations of Province A in 2024, the outpatient reimbursement rate for diabetic patients receiving treatment at designated hospitals is 70%", after preprocessing, key entities such as "diabetes", "70%", and "Province A" can be quickly located. The combination of entity recognition and relation extraction models significantly improves the accuracy of extracting the "disease type-ratio-region" association relationship, increasing efficiency by 8 times compared to manual extraction, avoiding omissions or mismatches caused by manual screening. The association relationship is transformed into tabular regulatory rules containing fields such as "disease type ID" and "region code", transforming policy logic from natural language into structured data that can be directly read by computers. When querying "Reimbursement ratio for diabetes in Province A", the system can directly locate the corresponding field in the table using "Region Code 370000 + Disease ID 001", reducing the response time to within 0.5 seconds. This solves the inefficiency of traditional text policy "word-by-word search" and provides high-speed data support for subsequent reimbursement review. A hash table structure model using region code and disease ID as the joint primary key achieves accurate mapping between "region-disease" and "reimbursement ratio". For example, inputting "Province A (370000) + Diabetes (001)" will instantly return a reimbursement ratio of "70%", improving query efficiency by more than 3 times compared to traditional database retrieval. This structure also supports multi-dimensional combined queries, meeting the needs of complex reimbursement scenarios involving multiple regions and diseases. When receiving local policy adjustment text (such as "Reimbursement ratio for diabetes in Province A adjusted to 75%)", natural language processing technology can automatically identify the adjustment object, the new ratio, and the effective time, without requiring manual modification of model parameters. The system automatically switches data upon the effective date (e.g., July 1, 2024), ensuring "zero delay" in policy implementation and significantly improving timeliness compared to manual updates (which average lag of 3 days). Simultaneously, the automatic triggering mechanism for the effective date avoids issues of "early implementation" or "delayed implementation," allowing patients to promptly enjoy policy benefits. The structured model and automatic update mechanism reduce manual intervention. For example, when seeking medical treatment across regions, the system automatically matches the corresponding proportions through region codes, avoiding errors in proportion application caused by manual judgment of region affiliation (e.g., applying a policy from province A to a patient in province B), thus significantly reducing the cross-regional reimbursement error rate. Furthermore, automatic model maintenance reduces repetitive work for policy administrators, allowing them to focus on core tasks such as policy interpretation, significantly lowering management costs.

[0028] In one embodiment, the material integrity verification in step S3 specifically includes: S301. Pre-set a material verification checklist according to the type of chronic and special diseases. For example, the checklist for diabetic patients includes "electronic medical records for the past 3 months" and "blood glucose test reports". S302. Perform OCR recognition on the received patient data to extract the material name and key information (such as the time of issuance of electronic medical record). S303. Compare the identification results with the verification list and mark the materials as "provided" or "missing". S304. Generate a "Material Missing List" for missing items, including the name of the missing material, supplementary requirements, and deadline; S305. Based on the data source marking, push the missing list to the corresponding terminal (such as the hospital HIS system, patient WeChat official account) and send SMS messages simultaneously; As described in steps S301-S305 above, a differentiated verification checklist is preset according to the type of chronic and special diseases, avoiding a "one-size-fits-all" approach to material requirements. For example, diabetic patients need to provide a "blood glucose test report," and hypertensive patients need to provide a "blood pressure test record," ensuring that the collected materials are highly matched with the needs of the disease, reducing the burden on patients to submit irrelevant materials, and avoiding omissions due to vague material requirements. The targeted checklist significantly improves the relevance of the submitted materials and reduces some invalid materials compared to the general checklist. OCR recognition technology is used to extract the material names and key information (such as the issuance time of electronic medical records) from patient data, replacing the traditional manual verification method. For example, it automatically identifies "electronic medical record in May 2024" and extracts the issuance time, quickly determining whether it meets the "last 3 months" requirement, significantly improving the recognition accuracy. The automatic comparison function with the verification checklist reduces the verification time for a single case from 5 minutes to 30 seconds, improving the overall verification efficiency by 8 times and avoiding omissions due to negligence during manual verification. The "Missing Materials List" not only clearly lists the missing items but also includes supplementary requirements (such as "a blood glucose test report issued by a hospital at or above level II is required") and deadlines, allowing patients or data submitters to clearly understand the required supplementary content and timeframes, reducing the cost of repeated communication. For example, if a diabetic patient is missing a "blood glucose test report from the past month," the list can directly suggest "a test report (including test values) after June 1, 2024 is required," significantly improving the first-time pass rate of supplementary materials. The missing materials list is pushed to the corresponding terminals based on the data source markings. For example, if data submitted through the hospital's HIS system is missing, it is directly pushed to the hospital's review interface; if data submitted by the patient is missing, it is simultaneously pushed to the official WeChat account and SMS. This targeted push model avoids mis-sending information, while SMS reminders ensure that recipients are promptly informed—significantly improving the feedback reach rate for missing materials and shortening the average supplementary response time from 2 days to 6 hours. By verifying the completeness of materials in advance and promptly reporting missing information, patient data is prevented from being returned due to incomplete materials after entering the subsequent review process, reducing the repetitive steps of "review-return-supplementation-re-review." Data shows that this process significantly reduces the review delay rate caused by missing materials, greatly improves the smoothness of the review process, and saves reviewers some ineffective review time.

[0029] In one embodiment, step S4, in which the artificial intelligence module parses medical records and determines the disease status, specifically includes: S401. Use a word segmentation tool to split the electronic medical record text into a word sequence, and use a pre-trained medical language model to identify keywords related to the disease state, such as "recovered" and "newly diagnosed hypertension"; S402. Combine the semantic context to eliminate ambiguity and distinguish between expressions such as "the patient has no symptoms of diabetes" (not cured) and "the patient's diabetes has been cured" (cured); S403. Compare the extracted keywords with historical medical data to determine the disease status (cured / existing / newly added disease). S404. Extract the insured location information from the patient data and convert it into a region code. Combine the disease ID with the reimbursement association model to obtain the reimbursement ratio. S405. Call the structured regulatory rules to verify the reimbursement conditions, comprehensively determine the reimbursement eligibility and generate the audit results; As described in steps S401-S405 above, using word segmentation tools combined with a pre-trained medical language model to identify disease status keywords enables precise location of core information from complex electronic medical record texts. For example, from the statement "The patient has a history of hypertension for 3 years, and a new diagnosis of hyperlipidemia has been added during this follow-up visit," the key phrase "newly added hyperlipidemia" can be accurately extracted, significantly improving keyword recognition accuracy. Compared to manually reading medical records sentence by sentence, this method not only increases information extraction efficiency by 10 times but also avoids omissions caused by misunderstandings of professional terminology during manual extraction (such as mistakenly ignoring the phrase "newly added"). Contextual semantic analysis eliminates ambiguity and effectively distinguishes easily confused medical record statements. For example, regarding "The patient has no symptoms of diabetes" and "The patient's diabetes has been cured," the system can accurately determine, based on contextual information such as "recent blood glucose tests are normal" and "no medication records," that the former indicates "current symptom relief" rather than cure, while the latter indicates "clinical cure," significantly improving the accuracy of ambiguity determination. This capability solves the problem of misjudging disease status due to semantic misunderstandings in traditional manual review, reducing some basic judgment errors. The logic of comparing keywords with historical medical data makes disease status determination more objective and continuous. For example, when determining "recovery," it not only relies on the keyword "recovery" in the current medical record but also requires historical data support of "no related medical records in the past 6 months." When determining "additional new disease," it is necessary to verify that there is no corresponding disease information in the historical records to avoid misjudging "relapse of an old disease" as "new disease." This multi-dimensional verification significantly improves the accuracy of disease status determination compared to traditional manual methods, providing a reliable basis for subsequent reimbursement eligibility determination. By converting the insured's place of insurance information into a regional code and combining it with the disease ID to call the reimbursement association model, the system achieves automated matching of reimbursement ratios. For example, for a diabetic patient whose insured's place of insurance is "Province A, City A," the system automatically converts the region into the code "370100" and quickly retrieves the corresponding 75% reimbursement ratio by combining it with the disease ID "001." The matching process takes less than 1 second and avoids the memory bias that occurs when manually querying regional policies (such as mistakenly remembering that the reimbursement ratio for diabetes in Province A is 70%). The cross-regional reimbursement ratio matching error rate has been reduced to below 3%. The fully automated process, from extracting keywords from medical records to generating final review results, has significantly shortened the reimbursement eligibility review cycle—reducing the review time per case from 15 minutes to less than 1 minute, and increasing the daily review volume by 15 times. Simultaneously, the automatic application of structured regulatory rules (such as verifying "whether treatment was received at a designated medical institution") ensures consistent review standards, avoiding the "different judgments for the same case" problem caused by differences in personal experience during manual review. This greatly improves the consistency of review results, enhancing both efficiency and fairness.

[0030] In one embodiment, the special patient identification and triage process in step S4 specifically includes: S406. Extract the total out-of-pocket expenses related to chronic and special diseases from the medical data for the past three months, and extract the average monthly income from the social security data; S407. The economic burden ratio is calculated by "out-of-pocket amount / (average monthly income × 3)". Patients with a ratio exceeding 30% are marked as having a high economic burden. S408. Extract family information from electronic medical records and community data, and identify special family circumstances such as "living alone" and "single-parent family" through a text classification model; S409. Score patients according to preset scoring criteria (adding 3 points for new diseases, 3 points for high economic burden, etc.) and generate level 1 / 2 / 3 special patient labels; S410: Match the corresponding priority reimbursement process based on the level label (Level 1 24-hour express process, etc.) and automatically assign review tasks; As described in steps S406-S410 above, the economic burden ratio is quantified using the formula "out-of-pocket amount / (average monthly income × 3)". Combined with medical data and social security income data from the past three months, this provides objective data support for identifying patients with a high economic burden. For example, if a patient's average monthly income is 5000 yuan and their out-of-pocket amount for chronic and special diseases in the past three months is 4500 yuan, the calculated ratio is 30%, and the system automatically marks them as a patient with a high economic burden. This data-driven identification method avoids biases caused by subjective experience in manual judgment, effectively improving the accuracy of identification and ensuring that patients who truly need economic support are accurately screened. Family information is extracted from electronic medical records and community data, and a text classification model is used to identify situations such as "living alone" and "single-parent families," overcoming the limitations of relying solely on patient self-reporting. For example, the system identifies information such as "patient lives alone, children live elsewhere" from community data, or extracts statements such as "single parent raising two minors" from electronic medical records, automatically including these patients in the special attention scope. This multi-source data integration approach significantly improves the identification coverage of special family circumstances, reducing omissions compared to traditional application methods and ensuring that assistance resources are tilted towards the most needy groups. Pre-set scoring criteria (e.g., 3 points for a new illness, 3 points for high economic burden) transform the abstract concept of "special circumstances" into quantifiable scores, avoiding the subjective arbitrariness of manual grading. For example, a single patient with a new illness and high economic burden living alone can obtain 3+3+2=8 points, generating a "Level 1" label; a patient with a single illness but high economic burden receives 3+1=4 points, generating a "Level 3" label. The transparent scoring logic allows patients to clearly understand the grading basis, while significantly improving the consistency of grading results and reducing disputes caused by ambiguous identification standards. Based on the level label, corresponding priority processes are matched, tilting resources towards patients with high needs. Level 1 patients enter a 24-hour expedited process, with review tasks automatically marked as "urgent" and prioritized for allocation to reviewers, ensuring settlement is completed in the shortest possible time; Level 2 patients complete their review within 48 hours, avoiding excessive waiting for patients with moderate needs. This differentiated processing model shortens the review and settlement cycle for Level 1 special patients from the traditional 3-5 days to within 1 day, and for Level 2 patients to within 2 days. This effectively improves the overall reimbursement efficiency for special patients and alleviates their financial burden. The system automatically allocates review tasks based on patient level labels, avoiding the inefficient "queueing" model where all patients wait in line. For example, Level 1 patients' review tasks skip the regular queue and are given priority access to review resources; ordinary patients are processed according to the standard procedure. This dynamic resource allocation method effectively improves the work efficiency of reviewers, reduces the contradiction of "high-demand patients waiting while low-demand patients occupy resources" caused by unreasonable resource allocation, and maximizes the effectiveness of limited review resources.

[0031] In one embodiment, the high-economic-burden patient assistance support in step S4 specifically includes: S411. From the system's built-in aid resource database, filter aid projects for the patient's specific disease and extract information such as application conditions and required materials; S412. Reuse medical insurance reimbursement materials to generate assistance application templates, automatically fill in existing information such as name and disease diagnosis certificate, and mark the content that needs to be supplemented; S413. Push the assistance application template and instructions to the patient's associated terminal (such as mobile APP). S414. Obtain the application progress status in real time through the information interface with the relief organization; S415. Record the progress information in chronological order to the "Rescue Records" subset of the patient's data file for the patient to query; As described in steps S411-S415 above, the system filters out assistance programs for patients' specific diseases from its built-in assistance resource library, avoiding the problem of patients blindly searching for assistance channels. For example, for diabetic patients with high economic burden, the system automatically matches them with special programs such as the "Diabetes Patient Charitable Assistance Program" and "Chronic and Special Disease Medical Assistance," and extracts application conditions such as "family monthly income is less than 1.5 times the local minimum standard" and required materials such as "income certificate and medical expense list." This precise matching eliminates the need for patients to search for assistance information one by one, significantly improving the matching rate of assistance programs and reducing most invalid applications compared to the patient self-searching model. The method of reusing medical insurance reimbursement materials to generate assistance application templates avoids the burden of patients repeatedly submitting materials. For example, materials such as "disease diagnosis certificate" and "medical expense details" already provided during medical insurance reimbursement are automatically filled into the assistance application template. Patients only need to supplement new content such as "income information of other family members," and the template will clearly indicate "needs to be supplemented." This material reuse model shortens the preparation time for assistance application materials from 3 days to 1 day, greatly improving the completeness of application materials and reducing application rejections due to missing materials. The system pushes application templates and guidelines for assistance to patients' associated devices (such as mobile apps), coupled with terminal message reminders, ensuring patients receive assistance information immediately. Compared to traditional "offline notifications" or "website announcements," this targeted push significantly improves reach and prevents missed opportunities due to delayed information access. For example, after a patient's medical insurance reimbursement application is approved, they immediately receive an assistance application notification on the app, achieving a seamless "reimbursement-assistance" process. Real-time progress updates (such as "application submitted," "under review," "assistance funds disbursed") are obtained through information interfaces with assistance organizations, addressing patients' anxiety about "no feedback after application." The system records progress information chronologically, allowing patients to check at any time via the app without repeatedly calling assistance organizations. This transparent progress tracking significantly improves patients' awareness of the assistance process, reduces most ineffective inquiries, and simultaneously forces assistance organizations to improve their review efficiency. A subset of "assistance records" stores application progress and results chronologically, providing historical evidence for subsequent assistance for patients. For example, patients whose initial applications are rejected can review the records to see why they were rejected "due to incomplete income documentation," allowing them to provide more targeted supplementary materials later. If the application is approved, the records serve as proof of "having received assistance," facilitating the assessment of future support needs by aid organizations. This archival management creates a closed loop for aid services, significantly increasing the success rate of subsequent aid applications and providing data support for precise assistance.

[0032] In one embodiment, step S5, manual review and result generation, specifically includes: S501, the manual review portal displays the review results of reimbursement eligibility and patient data in a visual interface, with the judgment result highlighted on the left and the material viewing entry provided on the right; S502. Supports reviewers to annotate comments online through the interface, and allows them to select commonly used options and add supplementary explanations; S503. Upon approval, the system generates a final reimbursement result containing the patient's basic information, reimbursement ratio, etc., and automatically affixes an electronic signature. S504. When the review fails, the reviewer selects the reason category and fills in the specific explanation to generate the final reimbursement result containing the reason for failure. S505. Store the final reimbursement result in a subset of the patient's data, including the "reimbursement records," and establish a unique association with the patient's identifier. As described in steps S501-S505 above, the visual interface prominently displays the judgment result on the left and provides a material viewing entry on the right, allowing reviewers to grasp the core information without repeatedly switching between multiple documents. For example, reviewers can directly see "Disease Status: Existing, Recommended Reimbursement Ratio: 75%" on the left, and click the entry on the right to view the corresponding original electronic medical record, without having to manually search for patient data. This information integration model reduces the review time per case from 15 minutes to 5 minutes, significantly improving review efficiency and avoiding omissions caused by scattered information. The online annotation function provides commonly used options and supplementary explanation boxes, transforming review opinions from "verbal statements" or "scattered records" into structured text. For example, selecting "Agree with AI review result" and adding "Materials complete, in compliance with Province A's policy," or selecting "Disease status does not match" and explaining the specific reasons, avoids subsequent disputes caused by ambiguous opinions. The standardized recording of review opinions improves the traceability of the review process to 100%, allowing for direct retrieval of the original opinions during subsequent reviews, reducing most of the repetitive communication caused by unclear opinions. When the review is approved, a result containing core information is automatically generated and an electronic signature is affixed, ensuring a uniform format and legal validity. When the review fails, the patient is required to select a reason category and fill in an explanation to avoid "unjustified rejections." For example, a clear record such as "Reason for rejection: The electronic medical record shows that the patient has recovered, which is inconsistent with the AI's determination of the continued existence status" not only makes the patient clearly aware of the basis for rejection but also provides a judgment standard for subsequent appeals. This standardized result generation model significantly improves the compliance rate of reimbursement results and reduces most disputes caused by non-standard results. The final result is stored in a subset of "reimbursement records" and linked to the patient's unique identifier, creating a complete archive of patient data and reimbursement results. For example, the patient's historical records such as "Reimbursement approved in July 2024 (75% rate)" and "Rejected in August 2024 due to missing materials" can be directly retrieved using the ID number, preventing the results from being out of sync with patient information. This associated storage model greatly improves data query efficiency and provides convenience for medical insurance departments to conduct statistical analysis and for patients to query historical reimbursement records, achieving refined management of "one person, one file; one report, one record." Human review serves as a supplement to AI review, correcting potential biases in AI (such as misjudging complex medical records), while a structured result generation process avoids the arbitrariness of human review. For example, if AI determines the record is "in existence" but a human reviewer finds the electronic medical record clearly shows "recovery," a standardized process can generate a rejection result, ensuring the final result is both efficient and accurate.

[0033] Example 2, please refer to Figure 2 As shown in this embodiment, a smart service system for chronic and special diseases includes: Policy and Rules Module: Configured to receive national and local policies input manually, transform them into structured regulatory rules through natural language processing technology, and establish and update the reimbursement association model; Data receiving and marking module: configured to receive patient data and mark the data source, and transmit patient data to the material verification module; The materials verification module is configured to analyze the integrity of patient data, transmit complete patient data to the AI ​​review module, and transmit incomplete patient data to the data source and indicate the missing data information. The AI ​​review module is configured to call the reimbursement association model and regulatory rules to review complete patient data, analyze patient disease information and determine reimbursement eligibility, and generate reimbursement eligibility review results. The manual review module is configured to receive the reimbursement eligibility review results and patient data, generate the final reimbursement result after manual review and confirmation, and update the patient's reimbursement eligibility status based on the final reimbursement result. Results Feedback Module: Configured to send the final reimbursement result back to the data source and generate a notification message when the final reimbursement result differs from the original reimbursement eligibility.

[0034] In a preferred embodiment, the policy rules module includes a policy input submodule, a rule parsing submodule, and an association model submodule, wherein: Policy Input Submodule: Configured to provide a manual input interface, supporting the uploading, modification, and storage of policy texts; The rule parsing submodule is configured to extract region-disease-reimbursement ratio information using natural language processing technology and convert it into computer-executable logic. The association model submodule is configured to build a region-disease-reimbursement ratio association model based on the extracted information and automatically update the model when policies are adjusted.

[0035] In a preferred embodiment, the policy rules module includes: The AI-powered review module includes a disease information analysis submodule, a reimbursement ratio calculation submodule, and a result output submodule. The disease information analysis submodule is configured to parse electronic medical records using natural language processing technology to determine the patient's disease status. The reimbursement eligibility determination submodule is configured to call the reimbursement association model and regulatory rules, and determine the patient's reimbursement eligibility based on the disease status and region and the matching reimbursement ratio. The results generation submodule is configured to generate the results of the reimbursement eligibility review and link them to the patient data. The special patient service module includes an identification and triage submodule, a reimbursement process adaptation submodule, and an assistance guidance submodule: The identification and classification submodule is configured to calculate the ratio of out-of-pocket expenses to average monthly income based on patient data, and generate special patient classification labels by combining special family circumstances. Reimbursement process adaptation submodule: configured to match the corresponding priority reimbursement process based on the level label; The assistance guidance submodule is configured to push assistance application guidance to patients with high economic burden, generate assistance application templates, and track progress.

[0036] The policy and rules module automates the entire policy process from input to application through a division of labor among sub-modules: "policy input," "rule parsing," and "association model." The policy input sub-module provides a standardized input interface, avoiding inconsistent policy text formats; the rule parsing sub-module uses natural language processing technology to extract core information, transforming "region-disease-reimbursement ratio" into computer-executable logic, significantly improving extraction accuracy; the association model sub-module automatically updates the model when policies are adjusted, reducing the time required for local policy changes from "3-5 days for manual modification" to "automatic system updates completed within 2 hours." The synergy of these three modules significantly improves policy implementation efficiency, ensuring patients can promptly enjoy policy benefits. The data receiving and marking module marks the data source, providing a basis for subsequent traceability; the material verification module analyzes the data integrity in a targeted manner, and pushes supplementary prompts for missing materials to avoid invalid transfers. For example, if a patient submits data through the hospital's HIS system and the electronic medical record is missing, the module will directly provide feedback to the HIS system requesting supplementation, significantly improving the first-time pass rate of supplementary materials. The two modules form a closed loop of "receiving-verification-feedback," reducing review delays caused by material issues and significantly improving the smoothness of data processing. The AI-powered review module has clearly defined sub-modules: the disease information analysis sub-module parses electronic medical records to determine disease status, avoiding human interpretation bias; the reimbursement eligibility determination sub-module accurately matches reimbursement ratios and conditions, reducing regional or disease mismatches; and the results generation sub-module links patient data to ensure traceability. The special patient service module provides differentiated services for patients with high economic burdens through identification and classification, process adaptation, and assistance guidance. For example, the review cycle for Level 1 special patients has been shortened from 3 days to 1 day. Overall, this significantly improves review efficiency and eligibility accuracy. The manual review module receives the AI ​​review results and patient data. Reviewers view the information and add comments through a structured interface to avoid ambiguity in their comments. After the results are generated, they are automatically linked to the patient data to form a complete record of "review-confirmation-storage". This model leverages the role of human error correction in complex cases while reducing human arbitrariness through standardized processes, ultimately significantly improving the compliance rate of reimbursement results. The results feedback module transmits the final reimbursement outcome back to the data source, allowing patients or hospitals to stay informed of the progress in real time. It automatically generates notifications when reimbursement eligibility changes, preventing patients from passively waiting. For example, when a patient's reimbursement eligibility changes from "disapproved" to "approved," they will receive an SMS notification containing the reason and the percentage, significantly improving information reach. This proactive feedback mechanism reduces the number of patient inquiries and significantly improves service satisfaction.

[0037] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A smart service method for chronic and special diseases, characterized in that, include: Receive manually input policy texts, which include national mandatory policy medical information and / or local policy medical information; The policy text was transformed into structured regulatory rules using natural language processing technology, and a reimbursement association model was established. Receive patient data and mark the data source, and analyze the completeness of the patient data; if the patient data is complete, transmit the complete patient data to the artificial intelligence module for review; if the patient data is incomplete, transmit the incomplete patient data to the data source and indicate the missing data information. The artificial intelligence module calls the reimbursement association model and the regulatory rules to review the complete patient data, analyze the patient's disease information, and determine the patient's reimbursement eligibility based on the disease information, and generate the reimbursement eligibility review result; The reimbursement eligibility results and the patient data are transmitted to the manual review port. After manual review and confirmation, the final reimbursement result is generated, and the patient's reimbursement eligibility status is updated according to the final reimbursement result. The final reimbursement result is transmitted back to the data source; and a notification message is generated when the final reimbursement result differs from the original reimbursement eligibility.

2. The intelligent service method for chronic and special diseases according to claim 1, characterized in that, After receiving manually input policy texts, a policy rule base is established, storing the policy texts according to the national policy-regional policy hierarchy. Key information such as disease name, reimbursement ratio, and applicable region is extracted from the policy texts using natural language processing technology and transformed into structured regulatory rules. Based on the extracted key information, a reimbursement correlation model of region-disease-reimbursement ratio is established. When manually input local policy adjustment texts are received, the adjustment information is located using natural language processing technology, and the reimbursement ratio correlation of the corresponding region is automatically updated.

3. The intelligent service method for chronic and special diseases according to claim 1, characterized in that, When the artificial intelligence module reviews complete patient data, it uses natural language processing technology to parse the electronic medical records in the complete patient data, extracts corresponding keywords, and compares historical treatment data to determine the patient's disease status; it calls the reimbursement association model and the regulatory rules to match the corresponding reimbursement ratio according to the disease status and the patient's region; based on the reimbursement ratio and disease status, it determines the patient's reimbursement eligibility and generates a reimbursement eligibility review result.

4. The intelligent service method for chronic and special diseases according to claim 3, characterized in that, When the artificial intelligence module reviews complete patient data, it calculates the economic burden ratio based on patient diagnosis and treatment data and integrated income data by dividing out-of-pocket amount by average monthly income. When the ratio reaches a high level, it is marked as a high economic burden. It also extracts special family circumstances by combining electronic medical records and community data to generate special patient level labels. The reimbursement process is matched to the corresponding priority based on the level label.

5. The intelligent service method for chronic and special diseases according to claim 4, characterized in that, For patients marked as having a high economic burden, the artificial intelligence module automatically pushes external assistance application guidelines, reuses medical insurance reimbursement materials to generate assistance application templates, and simultaneously tracks the assistance progress and records it in the patient's data file.

6. The intelligent service method for chronic and special diseases according to claim 1, characterized in that, The completeness of the analysis of the patient data includes: A pre-set material verification checklist is used to check the matching degree between the received patient data and the checklist through text recognition technology. Missing items are automatically generated into a "material missing list" and linked to the corresponding data source.

7. The intelligent service method for chronic and special diseases according to claim 1, characterized in that, The process of generating the final reimbursement result after manual review and confirmation specifically includes: The manual review portal simultaneously displays the reimbursement eligibility review results and corresponding patient data, allowing reviewers to annotate their comments online. When the review is approved, a final reimbursement result is generated. When the review is rejected, the reason for rejection must be filled in, and a final reimbursement result for rejection is generated. The final reimbursement result is synchronously linked to the patient data.

8. A smart service system for chronic and special diseases, used to implement the smart service method for chronic and special diseases as described in any one of claims 1-7, characterized in that, include: Policy and Rules Module: Configured to receive national and local policies input manually, transform them into structured regulatory rules through natural language processing technology, and establish and update the reimbursement association model; Data receiving and marking module: configured to receive patient data and mark the data source, and transmit patient data to the material verification module; The materials verification module is configured to analyze the integrity of patient data, transmit complete patient data to the AI ​​review module, and transmit incomplete patient data to the data source and indicate the missing data information. The AI ​​review module is configured to call the reimbursement association model and regulatory rules to review complete patient data, analyze patient disease information and determine reimbursement eligibility, and generate reimbursement eligibility review results. The manual review module is configured to receive the reimbursement eligibility review results and patient data, generate the final reimbursement result after manual review and confirmation, and update the patient's reimbursement eligibility status based on the final reimbursement result. Results Feedback Module: Configured to send the final reimbursement result back to the data source and generate a notification message when the final reimbursement result differs from the original reimbursement eligibility.

9. A smart service system for chronic and special diseases according to claim 8, characterized in that, The policy rules module includes a policy input submodule, a rule parsing submodule, and an association model submodule, wherein: Policy Input Submodule: Configured to provide a manual input interface, supporting the uploading, modification, and storage of policy texts; The rule parsing submodule is configured to extract region-disease-reimbursement ratio information using natural language processing technology and convert it into computer-executable logic. The association model submodule is configured to build a region-disease-reimbursement ratio association model based on the extracted information and automatically update the model when policies are adjusted.

10. A smart service system for chronic and special diseases according to claim 8, characterized in that, include: The AI-powered review module includes a disease information analysis submodule, a reimbursement ratio calculation submodule, and a result output submodule, as well as a special patient service module. The disease information analysis submodule is configured to parse electronic medical records using natural language processing technology to determine the patient's disease status. The reimbursement eligibility determination submodule is configured to call the reimbursement association model and regulatory rules, and determine the patient's reimbursement eligibility based on the disease status and region matching the reimbursement ratio. The results generation submodule is configured to generate the results of the reimbursement eligibility review and link them to the patient data. The special patient service module includes an identification and triage submodule, a reimbursement process adaptation submodule, and an assistance guidance submodule: The identification and classification submodule is configured to calculate the ratio of out-of-pocket expenses to average monthly income based on patient data, and generate special patient classification labels by combining special family circumstances. Reimbursement process adaptation submodule: configured to match the corresponding priority reimbursement process based on the level label; The assistance guidance submodule is configured to push assistance application guidance to patients with high economic burden, generate assistance application templates, and track progress.

Images