Inflammatory bowel disease home biologic injection adherence and adverse reaction early warning system

Abstract

The present application relates to the field of medical software, and discloses an inflammatory bowel disease home biological preparation injection compliance and adverse reaction early warning system, comprising an adverse reaction monitoring module configured to receive adverse reaction symptom feedback data and vital sign data uploaded by a user terminal mobile terminal at a preset time node, wherein the adverse reaction symptom feedback data comprises text description data and injection site image data; a natural language processing algorithm is used to extract symptom keywords from the text description data and match them with a preset adverse reaction symptom library, and an image recognition algorithm is used to extract local reaction feature parameters in the injection site image data; based on the matching results of the symptom keywords, the local reaction feature parameters and the abnormality level of the vital sign data, the adverse reaction risk level is determined as low risk, medium risk or high risk. The present application triggers multi-level early warning for low compliance and high-risk adverse reactions, and realizes whole-process closed-loop management and control of home injection.

Description

Technical Field

[0001] This invention relates to the field of medical software, and more particularly to a home-based biological agent injection compliance and adverse reaction early warning system for inflammatory bowel disease. Background Technology

[0002] Currently, novel biological agents used domestically and internationally for the treatment of inflammatory bowel disease (IBD) mainly include anti-adhesion molecules, anti-interleukin-12 / 23 agents, and JAK inhibitors. These novel biological agents offer advantages such as good efficacy, high safety, and convenient administration. In recent years, relevant departments in my country have successively approved several novel biological agents for the treatment of IBD, and relevant domestic and international guidelines also recommend them for first-line or second-line treatment of IBD with a high level of evidence.

[0003] One of the main advantages of most current biologic therapies is that patients can administer the medication subcutaneously at home using pre-filled syringes or pens. However, existing technologies lack a unified data platform, resulting in fragmented data that hinders clinical decision optimization, a lack of standardized home injection guidelines, and difficulties in patient compliance, adverse reaction monitoring, and early warning.

[0004] It can be seen that existing technologies lack a closed-loop management system that can organically integrate real-time monitoring of home injection behavior, quantitative assessment of compliance, intelligent identification and graded early warning of adverse reactions, standardized injection operation guidance, unified management of data throughout the entire process, and remote intervention by medical staff. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a home-based biological agent injection compliance and adverse reaction early warning system for inflammatory bowel disease, characterized in that it includes: a cloud data processing platform, and a user-end mobile terminal and a medical care-end management terminal respectively connected to the cloud data processing platform. The cloud-based data processing platform includes an intelligent injection assistance module, a compliance management module, an adverse reaction monitoring module, a multi-level early warning push module, and a data aggregation and analysis module. The intelligent injection assistance module is configured to break down the injection process of biological agents into multiple consecutive steps, send operation guidance data for the current step to the user mobile terminal, and after receiving the completion confirmation instruction for the current step from the user mobile terminal, send operation guidance data for the next sequential step to the user mobile terminal. The compliance management module is configured to send injection reminder instructions to the user's mobile terminal according to a preset injection cycle; receive the actual injection time and actual injection dose uploaded by the user's mobile terminal; calculate the deviation duration between the actual injection time and the predetermined injection time, and the dose deviation value between the actual injection dose and the predetermined single dose; and record the number of missed injections, the deviation duration, and the dose deviation value to generate a medication log. The adverse reaction monitoring module is configured to receive adverse reaction symptom feedback data and vital sign data uploaded by the user's mobile terminal at preset time nodes. The adverse reaction symptom feedback data includes text description data and injection site image data. Natural language processing algorithms are used to extract symptom keywords from the text description data and match them with a preset adverse reaction symptom database. Image recognition algorithms are used to extract local reaction feature parameters from the injection site image data. Based on the symptom keyword matching results, the local reaction feature parameters, and the abnormality level of the vital sign data, the adverse reaction risk level is determined to be low risk, medium risk, or high risk. The multi-level early warning push module is configured to use the number of missed injections, the duration of deviation, the dose deviation value, and the adverse reaction risk level as conditional variables to execute early warning judgment logic; when the low-level early warning triggering conditions are met, a low-level early warning instruction is sent to the user mobile terminal; when the medium-level early warning triggering conditions are met, a medium-level early warning instruction is sent to both the user mobile terminal and the medical staff management terminal; when the high-level early warning triggering conditions are met, a high-level early warning instruction is sent to the medical staff management terminal. The data aggregation and analysis module is configured to aggregate the medication log and the adverse reaction risk level according to a set time period, and extract and store the calculated number of missed injections, injection timeliness rate parameters and frequency of adverse reactions in the database.

[0006] Compared with the prior art, the beneficial effects of the present invention are as follows: By establishing individualized medication records for patients through a cloud platform, and relying on the user terminal to collect data on patients' home injection procedures, physical symptoms and signs, medication adherence is graded and assessed using deep learning algorithms. At the same time, adverse reaction symptoms are intelligently identified and risk-stratified. Multi-level warnings are triggered for low adherence and high-risk adverse reactions, and are pushed to patients and medical staff simultaneously. The entire process of data is standardized, aggregated and statistically analyzed to achieve closed-loop management of home injection. Attached Figure Description

[0007] Figure 1 This is a system structure diagram provided for an embodiment of the present invention. Detailed Implementation

[0008] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” or “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0009] A home-based biological agent injection adherence and adverse reaction early warning system for inflammatory bowel disease includes: A cloud-based data processing platform, and user-end mobile terminals and medical staff management terminals that are respectively connected to the cloud-based data processing platform; The cloud-based data processing platform includes an intelligent injection assistance module, a compliance management module, an adverse reaction monitoring module, a multi-level early warning push module, and a data aggregation and analysis module. The intelligent injection assistance module is configured to break down the injection process of biological agents into multiple consecutive steps, send operation guidance data for the current step to the user's mobile terminal, and send operation guidance data for the next step to the user's mobile terminal after receiving the completion confirmation instruction for the current step from the user's mobile terminal. The compliance management module is configured to send injection reminder instructions to the user's mobile terminal according to the preset injection cycle; receive the actual injection time and actual injection dose uploaded by the user's mobile terminal; calculate the deviation time between the actual injection time and the preset injection time, as well as the dose deviation value between the actual injection dose and the preset single dose; and record the number of missed injections, deviation duration, and dose deviation value to generate a medication log. The adverse reaction monitoring module is configured to receive adverse reaction symptom feedback data and vital sign data uploaded by the user's mobile terminal at preset time nodes. The adverse reaction symptom feedback data includes text description data and injection site image data. Natural language processing algorithms are used to extract symptom keywords from the text description data and match them with a preset adverse reaction symptom database. Image recognition algorithms are used to extract local reaction feature parameters from the injection site image data. Based on the symptom keyword matching results, local reaction feature parameters, and the abnormality level of the vital sign data, the adverse reaction risk level is determined to be low risk, medium risk, or high risk. The multi-level early warning push module is configured to use missed injections, deviation duration, dose deviation value, and adverse reaction risk level as conditional variables to execute early warning judgment logic; when the low-level early warning trigger conditions are met, a low-level early warning command is sent to the user's mobile terminal; when the medium-level early warning trigger conditions are met, a medium-level early warning command is sent to both the user's mobile terminal and the medical staff management terminal; when the high-level early warning trigger conditions are met, a high-level early warning command is sent to the medical staff management terminal. The data aggregation and analysis module is configured to aggregate medication logs and adverse reaction risk levels according to a set time period, and extract and store the calculated number of missed injections, injection timeliness rate parameters, and frequency of adverse reactions in the database.

[0010] In one implementation, the multiple step nodes, in the order of operation time, include the drug preparation node, site selection node, skin disinfection node, injection execution node, post-injection pressing node, and waste disposal node. The intelligent injection assistance module is also configured to respond to the node return command issued by the user's mobile terminal and reissue the operation guidance data of the completed step nodes before the current step node; after the intelligent injection assistance module receives a total of three completion confirmation commands for the complete injection process within the recorded continuous injection cycle, it grants the user's mobile terminal the permission to skip operation commands for each step node.

[0011] As one implementation method, the compliance control module executes the logic for sending injection reminder instructions and recording the number of missed injections as follows: One day and 30 minutes before the scheduled injection time, the first round of reminder instructions and the upcoming reminder instructions are sent to the user's mobile terminal, respectively. If the actual injection time is not received 8 hours after the scheduled injection time, an initial delay reminder instruction will be generated and sent. If the actual injection time is not received within 12 hours after the scheduled injection time, a reminder instruction for further delay will be generated and sent. If the actual injection time is not received within 24 hours after the scheduled injection time, the record will be marked as a missed injection, the number of missed injections will be accumulated, and a remote intervention notification will be sent to the medical staff management terminal.

[0012] As one implementation method, the preset time nodes are within 30 minutes, 24 hours and 72 hours after a single actual injection is completed. When the adverse reaction monitoring module determines that the preset time node has been reached and no adverse reaction feedback data has been obtained within the time limit, it triggers a pop-up selection command to the user's mobile terminal. Vital signs data include body temperature data, as well as a set of data consisting of resting heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate, and blood oxygen saturation; The adverse reaction monitoring module has built-in rules for classifying abnormal vital signs, which divide the acquired vital sign data into three levels: normal range, mild abnormality (Level 1), moderate abnormality (Level 2), and severe abnormality (Level 3). The adverse reaction monitoring module is configured to disable the abnormality judgment and comparison step for the missing data items when the received vital sign data includes body temperature data items but the other data items in the data item set are missing. Instead, it outputs the adverse reaction risk level based only on the abnormality level of the body temperature data items, the matching results of symptom keywords, and local reaction characteristic parameters.

[0013] As one implementation method, the adverse reaction monitoring module determines that the risk level of an adverse reaction is low under specific conditions by simultaneously satisfying the following set of elements: Local reaction characteristic parameters indicate that the erythema diameter is less than 5 cm and there are no induration features; The symptom keyword matching results do not contain words corresponding to the systemic symptom category; All vital signs data are within the normal range, or the number of data items in the first abnormality level is less than or equal to two and there are no data items in the second or third abnormality levels.

[0014] As one implementation method, the adverse reaction monitoring module determines the adverse reaction risk level as medium risk under specific conditions that any of the following conditions are met: Local reaction characteristic parameters indicate that the erythema diameter is greater than or equal to 5 cm and less than or equal to 10 cm, or the extraction result contains induration features; The symptom keyword matching results matched the mild systemic symptom classification thesaurus. The number of one or two data items in the vital signs data that are classified as the second abnormality level is 1; The number of data items in the vital signs data that are at the first abnormal level is greater than or equal to three.

[0015] As one implementation method, the adverse reaction monitoring module determines that the adverse reaction risk level is high under the specific condition that any one of the following conditions is met: Local reaction characteristic parameters characterize erythema diameter greater than 10 cm, or the extraction results include features of vesicles, ulceration or necrosis; The symptom keyword matching results matched the severe systemic symptom classification thesaurus. The vital signs data included data items at the third level of abnormality. The number of vital signs data items classified as the second abnormality level is greater than or equal to two; The adverse reaction monitoring module is configured with a high-risk direct trigger path: when the extracted symptom keywords match words in the preset emergency danger word library, or when large-area necrosis feature values ​​are extracted from the local reaction feature parameters, the adverse reaction monitoring module directly outputs a high-risk judgment result and suspends the abnormal level judgment logic thread for vital sign data.

[0016] As one implementation method, the warning triggering logic set by the multi-level warning push module includes: The triggering conditions for a low-level warning include any one of the following: the first single missed injection, a delayed record with a deviation duration greater than zero and less than or equal to 24 hours, a non-zero dose deviation value and the acquisition of input data containing corrective coverage actions, or an adverse reaction risk level of low risk. The triggering conditions for a medium-level warning include: the cumulative number of missed injections reaching two within two consecutive judgment periods, the deviation duration being greater than 24 hours but less than the preset latest supplementary injection time threshold, obtaining two consecutive non-zero dose deviation values, or any one of the following: the adverse reaction risk level is medium risk.

[0017] As one implementation method, the warning triggering logic set in the multi-level warning push module also includes: The triggering conditions for a high-level warning include any of the following: the duration of a single missed injection is greater than the latest refill time threshold; missed injection status is recorded in three or more consecutive judgment periods; the duration of a single deviation is greater than the latest refill time threshold; a non-zero dose deviation value is obtained three or more times; or the adverse reaction risk level is high. The high-level warning command is configured to, upon being received by the medical staff management terminal, execute a forced pop-up interrupt command on the interface and activate the emergency voice dialing call interface.

[0018] As one implementation method, the data aggregation and analysis module is also configured as follows: Intervention record data, which includes communication timestamps, interaction method tags, and processing opinion text, is entered upon receiving medium- or high-level warning commands from the medical care management terminal. The intervention record data is mapped and bound to the medication logs and adverse reaction risk levels of the data source that triggered the corresponding warning commands based on foreign keys. Extract compliance quantitative scoring variables and classification distribution matrix sets of adverse reactions of different risk levels within a set historical time window. Calculate the correlation between missed injections and dosage error frequency and the distribution of adverse reaction risk levels. Update the classification feature set of the preset adverse reaction symptom database by calculating the correlation weights returned by the calculation, and modify the delay threshold tolerance parameters used to determine the triggering conditions for each level of warning accordingly.

[0019] Specifically, such as Figure 1This invention provides an intelligent early warning management system for home-based IBD biologic injection compliance and adverse reactions, comprising: 1. User-end mobile terminal The user-end mobile terminal (such as a mobile app or WeChat mini-program) is used by patients to enter their basic personal information, IBD condition data (disease type, lesion site, disease course, past medication history, etc.), biological agent medication regimen (drug name, single dose, injection cycle), collect home injection operation videos or pictures, adverse reaction symptom feedback data (text description, pictures or videos), vital sign data (body temperature, blood pressure, etc.), and receive medication reminders and warning information.

[0020] 2. Medical staff management terminal The healthcare management terminal allows healthcare professionals to view medication adherence data (medication logs, adherence scores and levels) and adverse reaction monitoring data (symptom records, risk level assessment results) of patients under their management. It also enables them to receive abnormal alerts, perform remote interventions (online consultations, telephone follow-ups), provide medication guidance, and manage follow-ups. Healthcare professionals can also view automatically generated periodic adherence reports and chronic disease management follow-up reports through this terminal.

[0021] 3. Cloud-based data processing platform The cloud-based data processing platform, serving as the system's core computing and storage center, is responsible for establishing and managing individualized patient medication records, aggregating and processing data from various modules, operating adherence assessment and adverse reaction identification models, and determining and distributing early warning logic. The platform also includes a data encryption and privacy protection unit to encrypt and store patient condition and medication privacy data during transmission, allowing only authorized medical personnel to access the corresponding patient data.

[0022] 4. Intelligent Injection Assist Module The intelligent injection assistance module incorporates standardized injection procedures for IBD biologics (including pre-filled syringes and pre-filled pens), providing real-time guidance, standardization verification, and error alerts for patients' home injection procedures. Specific functions include: injection site location (illustrated guidance for recommended sites such as the abdomen and thigh), disinfection procedure reminders (disinfection area, time, etc.), drug mixing detection (for drugs requiring reconstitution), injection completion confirmation, and reminders for post-injection pressure duration.

[0023] This module employs a step-by-step interactive video guidance approach: the complete injection procedure is broken down into multiple independent steps (e.g., "preparing medication → selecting the injection site → disinfecting the skin → performing the injection → post-injection pressure → waste disposal"), each corresponding to a short video or text description. After completing each step, the patient clicks "Next," at which point the system plays the next instruction. This design ensures that patients perform the procedures in the correct order, avoiding skipping steps or missing crucial steps, effectively reducing the risk of operational errors. After completing the video tutorial, patients can return to any previous step for repeated review, ensuring they have mastered the correct injection method.

[0024] 5. Compliance Management Module The compliance management module is connected to the cloud data processing platform and includes the following sub-units: (1) Medication Reminder Unit: Based on the patient's individualized biological agent injection cycle and time nodes, injection reminders are sent through various means such as SMS, in-app push, and voice calls. The system supports a multi-level reminder strategy—the first reminder is sent one day before the scheduled injection time, and the second reminder is sent 30 minutes before the injection; if the patient does not confirm that the injection has been completed, the system will trigger the first reminder 8 hours after the scheduled time; if there is still no response, the system will trigger the second reminder 12 hours later; if the patient has not completed the injection within 24 hours, the system will actively notify medical staff to intervene.

[0025] (2) Medication Recording Unit: Automatically records the time and dosage information of patient confirmation injection, missed injection, and delayed injection, and generates a detailed medication log. The medication log includes information such as the scheduled time of each injection, the actual injection time, the delay time, the injection dosage, the injection site, and whether the injection was completed.

[0026] (3) Compliance assessment unit: Based on medication log data, combined with missed injection rate, injection on-time rate (the deviation between the actual injection time and the scheduled time), and dosage accuracy (whether the dosage used by the patient matches the actual dosage required by the background), the patient's medication compliance is quantitatively assessed through a weighted scoring algorithm. A compliance report is automatically generated every cycle (such as monthly or quarterly) and pushed to the patient and medical staff.

[0027] 6. Adverse Reaction Monitoring Module The adverse reaction monitoring module includes the following sub-units: (1) Data Acquisition Unit: Automatically reminds patients to upload adverse reaction symptom feedback data according to preset time nodes (such as within 30 minutes, 24 hours, 72 hours after injection, etc.), and supports data upload in multiple formats such as text description, local photos, and videos. At the same time, it supports patients to actively report discomfort symptoms that occur at any time. Patients can simultaneously upload vital signs data such as body temperature, pulse / heart rate, blood pressure, respiration, and oxygen saturation.

[0028] (2) Risk Identification Unit: A database of common adverse reactions to IBD biological agents is set up, covering local adverse reactions (erythema, swelling, induration, pain, itching, rash, etc. at the injection site), systemic adverse reactions (fever, fatigue, headache, joint pain, gastrointestinal reactions, etc.), and serious adverse reactions (allergic reactions, dyspnea, chest pain, signs of severe infection, etc.). Semantic recognition methods are used to extract keywords and match patterns from patients' text descriptions of symptoms. Then, image recognition algorithms are used to analyze the range of redness and swelling, severity, and other features of the photos of the injection sites uploaded by patients to comprehensively determine the type of adverse reaction.

[0029] (3) Risk grading unit: Based on the severity, duration, impact on daily life and abnormal vital signs of adverse reaction symptoms, adverse reaction risk is divided into three levels: low risk, medium risk and high risk.

[0030] Low risk: Mild local reactions (such as slight erythema and mild pain) that do not affect daily activities and vital signs are normal.

[0031] Medium risk: Moderate local reaction (such as obvious redness and swelling, persistent pain) or mild systemic reaction (such as low-grade fever, fatigue), which has a certain impact on daily activities, but vital signs are basically normal.

[0032] High risk: severe local reaction (large area of ​​redness and swelling, severe pain) or obvious systemic reaction (high fever, chills, difficulty breathing, severe allergic signs), or significantly abnormal vital signs, or symptoms that continue to worsen.

[0033] 7. Multi-level early warning push module The multi-level early warning push module triggers corresponding level early warnings based on compliance classification results and adverse reaction risk levels, and simultaneously pushes the early warning information to both the user's mobile terminal and the medical staff's management terminal. The early warning mechanism is as follows: Low-level alert: The trigger condition is the patient experiencing any of the following situations for the first time: a single missed injection, a single delayed injection but the delay time does not exceed 24 hours, a single incorrect dose but which has been corrected in time, or the system determines it to be a low-risk adverse reaction. Only medication correction guidance, symptomatic care suggestions, or health education content are pushed to the user's mobile terminal; no alerts are triggered on the healthcare provider's terminal.

[0034] Medium-level warning: The trigger condition is that the patient experiences any of the following situations within two consecutive injection cycles: two cumulative missed injections, a single delayed injection exceeding 24 hours but not exceeding the latest re-injection time for the medication, two consecutive dosage errors, or the system determines it to be a medium-risk adverse reaction. Warning information is simultaneously pushed to both the user and healthcare provider's devices to remind healthcare staff to pay attention and provide remote communication and guidance.

[0035] High-level alert: Triggered when a patient experiences any of the following: a single missed injection exceeding the latest re-injection time, missed injections occurring in three or more consecutive injection cycles, a single delayed injection exceeding the latest re-injection time, three or more cumulative dosage errors, or the system determines it to be a high-risk adverse reaction. An emergency alert will be immediately sent to healthcare professionals (supporting multiple methods such as SMS, telephone, and in-app strong reminders), initiating a remote emergency intervention process, and, if necessary, guiding the patient to seek medical attention at the nearest facility.

[0036] 8. Data Summary and Analysis Module The data aggregation and analysis module interfaces with the cloud-based data processing platform, specifically responsible for the unified collection, organization, statistical analysis, and visualization of historical data collected throughout the system's lifecycle, including patient medication behavior, injection procedures, adverse reactions, and medical interventions. This provides intuitive data support for individualized management, clinical follow-up, and strategy optimization. This module includes the following sub-units: (1) Data integration and standardization unit: According to the patient ID, time period, drug type, adverse reaction type, etc., all data from the user end, medical staff end and various functional modules are automatically aggregated and stored in a dedicated chronic disease management database to ensure data integrity and traceability.

[0037] (2) Periodic Statistical Analysis Unit: Automatically calculates core management indicators for patient groups on a daily, weekly, monthly, and quarterly basis, such as missed injection rate, injection timeliness rate, compliance level distribution, adverse reaction incidence rate, and proportion of different risk levels. It supports longitudinal comparison of historical data for individual patients and horizontal comparative analysis of patient groups with different medication regimens.

[0038] (3) Automatic report generation unit: Based on the statistical analysis results, it automatically generates individualized medication adherence reports, adverse reaction monitoring reports and chronic disease management follow-up reports. It supports export and sharing in chart form, which makes it convenient for medical staff to intuitively understand the patient's condition and provides a basis for adjusting medication regimens, formulating follow-up plans and carrying out targeted patient education.

[0039] (4) Trend Analysis and Model Optimization Unit: Utilizing long-term accumulated real patient data, this unit explores potential correlations between medication adherence and disease recurrence risk, and between adverse reactions and medication regimens. Based on these patterns, it continuously optimizes the weight parameters of the adherence assessment algorithm, the feature library of the adverse reaction identification model, and the warning threshold settings, thereby continuously improving the system's intelligent judgment accuracy and management effectiveness.

[0040] Based on the above management system, the present invention also provides an intelligent early warning management method for IBD biologic injection compliance and adverse reactions at home via the internet, comprising the following steps: Step S1 (File Creation): S1-1. Patients access the system via mobile devices (mobile app or WeChat mini-program), complete real-name authentication and account registration. The system first retrieves information from the inpatient system; patients can also fill in information themselves. The content includes: Personal basic information: including name, gender, date of birth, height, weight, contact information, emergency contact person and telephone number, allergy history, etc.

[0041] Medical information includes: disease type: Crohn's disease / ulcerative colitis, diagnosis time, main lesion site, previous surgical history, previous medication history, etc.

[0042] S1-2. Enter the medication regimen: Enter the information of the currently used biological agents, including: Drug Name: The system provides a drop-down list covering common IBD biologics such as adalimumab, vedelizumab, ustekinumab, and gusejinumab.

[0043] Single injection dose; Injection cycle: once every two weeks or once every four weeks, forming a fixed cycle according to the biological agent used by the patient.

[0044] Injection scheduled date: The time is uniformly set to 9:00 AM on the injection date.

[0045] S1-3. Establishing a Personalized File: The system uploads the above information to the cloud data processing platform to create a unique personalized medication management file for each patient. Subsequently, all medication records, adverse reaction data, early warning information, and medical intervention records related to that patient are linked to this file, forming a complete personal management log.

[0046] Step S2 (Reminders and Instructions): S2-1. Initial Reminder: Based on the injection cycle set by the system, the system will send an initial reminder to the patient via App push and SMS one day before the scheduled injection date. The message will be something like: "[Medication Reminder] Your scheduled injection of biological agent is tomorrow. Please prepare in advance." S2-2. Approaching Reminder: 30 minutes before the scheduled injection time, the system will send another reminder, such as: "[Approaching Reminder] Your injection time is approaching. Please prepare your medication and injection supplies. Click to enter the injection instructions." S2-3. Entering the Injection Guidance Process: After the patient clicks the reminder notification, they will enter the step-by-step interactive guidance interface of the intelligent injection assistance module. This interface breaks down the entire injection process into the following 6 steps, each accompanied by a short video or text explanation: Step-by-step instructions ① Prepare medication: Check the medication name and expiration date. Take the medication out of the refrigerator and warm it up (if needed). Video demonstration: How to take out the pre-filled pen / syringe and check the color and expiration date of the medication.

[0047] ② Choose the injection site: Select the abdomen (5cm away from the navel) or the front of the thigh as the injection site, avoiding areas with scars and redness. The diagram shows the recommended areas and areas to be avoided. Patients can take photos of the injection sites and upload them for system verification.

[0048] ③ Disinfect the skin with alcohol wipes, using a spiral motion from the inside out around the injection point (disinfection range ≥ 5cm), and let it dry. The video demonstrates the correct disinfection technique, and the system prompts "Please wait for the alcohol to completely evaporate before injecting".

[0049] ④ To administer the injection, pinch the skin, insert the needle at the standard angle, slowly inject the medication, and then remove the needle. The video demonstrates the needle insertion angle (45° or 90°) and injection speed.

[0050] ⑤ After injection, press the injection site with a dry cotton ball for 1-2 minutes. Do not rub. The video demonstrates the pressing technique and reminds users "Do not rub to avoid leakage of the medication".

[0051] ⑥ Waste disposal: Dispose of used syringes / pen needles in the sharps container and provide instructions on safe disposal methods. The operation guide is for patients receiving their first 1-3 injections and requires them to watch it. Patients receiving subsequent long-term injections may choose to watch it selectively.

[0052] S2-4. Step-by-step confirmation mechanism: After completing each step, the patient must click the "Completed, Next Step" button on the interface before the system unlocks the next section of guidance. Patients can also click "Back" at any time to repeat the learning process. This design ensures that patients do not skip key operational steps, reducing the risk of injection failure or local reactions due to improper operation.

[0053] Step S3 (Recordation and Compliance Assessment): S3-1. Injection Completion Confirmation: After completing the injection, the patient clicks the "I have completed the injection" button on the mobile device. The system will display a confirmation page, requiring the patient to confirm the following information: Actual injection time (default is the current time, which the patient can manually modify). Actual injection dose (entered manually by the patient, or the patient can upload the injected medication; the system will match this dose with the "should be injected" record). Injection site (select from options: left abdomen, right abdomen, left thigh, right thigh, etc.) Have any unusual circumstances occurred? (Optional) S3-2. Handling Delayed Injection: If the patient does not confirm the injection within 2 hours of the scheduled injection time, the system will automatically trigger a second reminder: "You have not confirmed the injection. Please make up for any missed injections as soon as possible." If confirmation is still not made after 4 hours, the system will remind the patient again. If the patient does not make any confirmation within 24 hours, the system will mark the injection as "missed" and automatically notify the healthcare provider for intervention (contacting the patient via SMS, WeChat, or phone). If the latest injection time has passed, an emergency alert will be immediately sent to the healthcare provider (supporting multiple methods such as SMS, phone, and in-app strong reminders), initiating a remote emergency intervention process, and guiding the patient to seek medical attention at the nearest facility if necessary.

[0054] S3-3. Medication Log Generation: The medication recording unit automatically records information such as the scheduled time, actual injection time, delay duration, injection dosage, injection site, and completion status of each injection, generating a structured medication log, which is then stored on the cloud data processing platform.

[0055] S3-4. Adherence Data Update: Based on medication log data, the adherence assessment unit continuously calculates the following objective indicators: Missed injections: The total number of injections not performed within the statistical period; Number of delayed injections and duration of delay: Record the specific duration of each delay; Dosage accuracy: Compare the injection dose confirmed by the patient with the prescribed dose and record any discrepancies. These objective data will serve as the direct basis for subsequent early warning triggering and multi-level early warning determination. Each cycle (e.g., monthly) the system will automatically generate a compliance report, presented to the patient in a simple chart (e.g., "This month you had 3 injections on time, 1 delay, and 0 missed injections"), and simultaneously pushed to the healthcare staff for reference.

[0056] Step S4 (Adverse Reaction Monitoring): S4-1. Timed Reporting Reminders: The system automatically sends reporting reminders to patients based on preset adverse reaction monitoring time points. Within 30 minutes after injection: Focus on monitoring for acute allergic reactions. A reminder will be given: "It has been 30 minutes since the injection. Do you experience any discomfort such as dizziness, difficulty breathing, or itchy skin?" 24 hours after injection: Focus on monitoring the injection site reaction and early systemic reactions. Reminder: "It has been 24 hours since the last injection. Please check if there is redness, swelling, or pain at the injection site, or if there is fever, fatigue, or other discomfort." 72 hours after injection: A reminder to report again and pay attention to delayed adverse reactions.

[0057] (Home monitoring of delayed adverse reactions mainly occurs 1-3 days after injection, including delayed rash, persistent induration, fever, blisters, and herpes zoster-like pain.) If the report is not submitted by the deadline, the system will automatically send a reminder. If the patient still has not filled out the form after 72 hours, a pop-up window will automatically appear, allowing the patient to select whether they experienced any adverse reactions. If the patient still has not filled out the form after 24 hours, a pop-up window will appear on the healthcare system's interface to remind them to intervene (by contacting the patient via SMS, WeChat, or phone to inquire about any adverse reactions). S4-2. Symptom Reporting Methods: After receiving the alert, patients can report their physical reactions through the following methods: Quick Questionnaire: The system provides structured questionnaires such as "Is there pain at the injection site? (None / Mild / Significant / Severe)", "Is there redness and swelling on the skin? (None / Mild / Moderate / Severe)", and "Do you have a fever? (None / Below 37.5℃ / 37.5-38.5℃ / Above 38.5℃)". Patients can select their desired option.

[0058] Text description: Patients can freely describe their discomfort in the text box, such as "The injection site is a little itchy and feels a little hard to the touch".

[0059] Photo upload: Patients can take a photo of the injection site and upload it. The system prompts "Please take the photo in a well-lit environment and keep the phone about 15-20cm away from the skin".

[0060] Vital signs entry: Patient's temperature, pulse / heart rate, blood pressure, respiration, oxygen saturation, etc. are recorded; S4-3. Intelligent Risk Identification: After receiving the data reported by the patient, the risk identification unit immediately initiates dual intelligent analysis: Semantic analysis: Using natural language processing (NLP) algorithms, symptom keywords (such as "itch," "red," "swelling," "heat," and "dizziness") in the patient's text description are extracted and matched with the built-in adverse reaction symptom database.

[0061] Image analysis: Using image recognition algorithms, the injection site photos uploaded by patients are analyzed to extract features such as the area, color depth, and edge morphology of the red and swollen area, and to comprehensively assess the severity of the local reaction.

[0062] Vital signs: Data collection can be automated through wearable devices (such as smartwatches) or manually entered by the patient using home medical devices (such as electronic blood pressure monitors and thermometers). Body temperature is a mandatory field for patients and is a required reference. Other vital signs are not required reference conditions. The system does not automatically classify missing vital sign data as abnormal; instead, it makes a comprehensive judgment based on existing data (such as body temperature) combined with symptom descriptions. Specific reference standards are shown in Table 1: S4-4. Risk Level Determination: The risk grading unit automatically determines the risk level of the current adverse reaction by combining the results of semantic and image analysis. Low-risk injection sites may present with mild erythema, mild pain, and transient itching, without induration, and do not affect daily activities; there are no systemic symptoms and vital signs are normal. Only data is recorded, and self-care advice is provided to the patient (such as "local cold compresses can relieve discomfort, and it usually subsides on its own within 1-2 days"). The system comprehensively assesses symptoms (diameter of erythema, degree of pain, systemic reaction) with the number and severity of abnormal vital signs, according to the following rules: 1. Low risk – corresponds to mild adverse reactions A risk level is considered low if all of the following conditions are met: Symptoms: Erythema at the injection site <5cm in diameter, mild pain or itching, no induration; no systemic symptoms (such as fever, fatigue, headache, etc.). Vital signs: All vital signs are within the normal range, or only 1-2 signs are mildly abnormal (Grade I), and no single sign meets the criteria for moderate abnormality. Functional impact: Does not affect daily activities. System Response: Only records data and pushes self-care suggestions to the patient (e.g., "You can apply a cold compress to relieve discomfort, which usually subsides on its own within 1-2 days. If symptoms worsen or new discomfort occurs, please report it promptly."). 2. Medium risk – corresponding to moderate adverse reactions A risk level is determined to be medium if any of the following conditions are met: Symptoms: Erythema at the injection site with a diameter ≥5cm and ≤10cm, accompanied by persistent pain or induration Mild systemic symptoms may occur (such as low-grade fever of 37.3℃-38.0℃, mild fatigue, headache, etc.). Vital signs meet any of the following conditions: One or two items showed moderate abnormalities (Level II). Three or more mild abnormalities were observed (Grade I). Functional impact: Has some impact on daily activities. System Response: Record the data and mark it as "requiring attention," then push it synchronously to healthcare professionals. This reminds healthcare staff to contact the patient within 24 hours for remote assessment and intervention.

[0063] 3. High risk – corresponding to severe or worse adverse reactions A risk level is determined when any of the following conditions are met: Symptoms: Erythema at the injection site >10cm in diameter, severe pain, blisters, ulceration, or necrosis. Or obvious systemic symptoms may appear (high fever >38.5℃, chills, difficulty breathing, chest tightness, laryngeal edema, generalized rash and other signs of severe allergic reaction). Vital signs meet any of the following conditions: If any one of the following severe abnormalities (Grade III) is present (such as heart rate <40 beats / min, systolic blood pressure <80 mmHg, SpO2 <85%, etc., regardless of whether other indicators are normal), the result is considered abnormal. Two or more moderate abnormalities (Grade II) were found. When vital signs are extremely unstable or a certain indicator reaches the threshold for acute and critical illness. Functional impact: Significantly affects daily activities or prevents self-care. Special triggering mechanism: If the symptom data uploaded by the patient contains preset emergency keywords (such as "difficulty breathing", "shortness of breath", "throat tightness", "unconsciousness", "fainting", etc.), or if the system image recognition determines that there are large areas of necrosis at the injection site, it will be directly judged as high risk without waiting for complete vital sign data.

[0064] System Response: An emergency pop-up window is immediately triggered on the healthcare provider's device, and healthcare personnel are notified via SMS and automated voice call. Healthcare personnel immediately contact the patient for emergency intervention, and if necessary, guide the patient to the nearest emergency room. Simultaneously, an emergency guidance message is sent to the patient's device (e.g., "Your reported symptoms require immediate attention from healthcare personnel. Please keep your phone accessible. If your breathing difficulties worsen, please call 120 immediately."). S4-5. Proactive Reporting Function: In addition to timed reminders, patients can proactively report symptoms at any time they feel unwell via the "Report Discomfort" button on their device. The system uses the same intelligent analysis and grading process for proactively reported data as for timed reports, ensuring that any abnormalities are captured and processed promptly.

[0065] My step S5 (alarm triggered): S5-1. Data Aggregation: The cloud-based data processing platform aggregates medication adherence data (missed doses, delays, dosage deviations) generated in step S3 and adverse reaction risk level data generated in step S4 in real time.

[0066] S5-2. Warning Level Determination: The multi-level warning push module determines the warning level according to preset objective rules, as follows: Warning Level Triggering Conditions (any one of them will trigger the warning) Notification Recipients Notification Method: Low-level warnings include: ① First instance of a single missed injection; ② Single delayed injection with a delay duration of ≤24 hours; ③ Single dose input error that has been corrected; ④ Low-risk adverse reactions are only reported via patient-side notifications. Medium-level warning: ① Two missed injections within two consecutive injection cycles; ② A single injection delayed by more than 24 hours but not exceeding the latest time for re-injection; ③ Two consecutive dose input errors; ④ Medium-risk adverse reaction. Patient-side + Healthcare-side: Patient-side: Information push + SMS; Healthcare-side: System messages + to-do item marking; Reminding healthcare staff to intervene (SMS, WeChat, telephone contact with patient). High-level warnings: ① Single missed injection exceeding the latest re-injection time for the drug; ② Missed injections occurring in three or more consecutive injection cycles; ③ Single delayed injection exceeding the latest re-injection time; ④ Three or more cumulative dose input errors; ⑤ High-risk adverse reactions. Patient side + medical staff side (emergency): Patient side: push notification + SMS; Medical staff side: emergency pop-up + SMS + telephone.

[0067] S5-3. Warning Content Push: The content sent to patients consists of easy-to-understand health guidance information. For example, in the case of a low-risk adverse reaction, the message might be: "The mild pain at the injection site you reported is a common phenomenon and usually does not require special treatment; you can apply a cold compress locally. If the symptoms worsen, please report it promptly." The content pushed to healthcare professionals is a summary of the patient's status. For example, a medium-risk alert would display: "Patient Zhang XX, adalimumab treatment, reported redness and swelling at the injection site approximately x cm in diameter 24 hours after this injection, accompanied by mild pain. The system has determined this to be a medium-risk case, and remote follow-up within 24 hours is recommended." Step S6 (Remote Intervention and Closed Loop): S6-1. Viewing Alerts on the Healthcare Staff Terminal: Healthcare staff log in to the healthcare staff management terminal and can arrange all pending alerts in the "Pending Alerts" list on the homepage according to urgency level (high > medium > low) and time order. Clicking on any patient allows viewing that patient's complete file, including: Basic information and medication regimen, historical medication log and adherence trend chart, records of all adverse reactions reported and their risk levels, and details of this warning.

[0068] S6-2. Remote Intervention Implementation: Medical staff select appropriate intervention methods based on the warning level and the patient's specific condition. Online text and image consultation: Through the system's built-in instant messaging function, users can engage in text and image conversations with patients to understand their detailed situations and provide guidance. Conversation records are automatically archived to the patient's file.

[0069] Telephone follow-up: Communication is conducted via voice using the contact number in the patient's file. The system provides follow-up record templates, allowing medical staff to quickly select and fill in key communication points and conclusions.

[0070] Medication regimen adjustment suggestions: If severe adverse reactions occur repeatedly or compliance is extremely poor, medical staff can send medication regimen adjustment suggestions to senior doctors through the system. After review, the medication regimen in the patient's file will be updated.

[0071] S6-3. Intervention Result Recording: After completing the intervention, medical staff record the intervention time, method, key communication points, patient feedback, and follow-up treatment opinions in the system. This record is linked to the original warning, forming a complete "warning → intervention → feedback" closed loop.

[0072] S6-4. Referral Guidance: For patients identified as needing emergency medical attention during high-risk warnings, medical staff will guide them to the nearest medical institution through the system, informing them of the necessary documents and precautions.

[0073] Step S7 (Data Summarization and Report Generation): S7-1. Regular Data Summarization: The cloud-based data processing platform automatically summarizes and calculates all data quarterly / semi-annually, updating the following metrics: Individual indicators: cycle missed rate, average delay time, dose accuracy, frequency and grade distribution of adverse reactions for each patient.

[0074] Population indicators: average compliance and incidence of adverse reactions, grouped by drug type, patient disease duration, age group, etc.

[0075] S7-2. Individual Follow-up Report Generation: The system automatically generates a graphic-rich "Biologic Injection Management Follow-up Report" for each patient, including the following content: This report includes a summary of medication use during this period (planned number of injections, actual number of injections completed, and on-time completion rate), a line graph of adherence trends (changes over the last 6 months), a summary of adverse reaction records, and information on improvements or changes compared to the previous period. The report can be exported as a PDF for patients to present to their doctors during in-person visits.

[0076] S7-3. Group Analysis Report Generation: The data summary and analysis module generates group-level statistical reports for medical staff, supporting the following analytical perspectives: Comparison of compliance with different biologics; Correlation analysis of adverse reaction incidence with drug type and injection site selection; Correlation analysis between the frequency of early warning triggers and the effectiveness of medical interventions.

[0077] S7-4. Continuous Model Optimization: Based on the system's long-term accumulation of real-world data and trend analysis, the following models are periodically iteratively optimized: Compliance prediction model: Combining patients' historical behavioral characteristics to predict the risk of future missed injections and intervene in advance.

[0078] Adverse reaction identification model: Expand the feature library based on newly added images and symptom data to improve identification accuracy.

[0079] Early warning threshold optimization: Based on the actual response feedback from medical staff, the triggering conditions for each level of early warning are dynamically adjusted to balance the sensitivity and specificity of the early warning. The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A home-based biological agent injection compliance and adverse reaction early warning system for inflammatory bowel disease, characterized in that, include: A cloud-based data processing platform, and a user-end mobile terminal and a medical care management terminal that are respectively connected to the cloud-based data processing platform. The cloud-based data processing platform includes an intelligent injection assistance module, a compliance management module, an adverse reaction monitoring module, a multi-level early warning push module, and a data aggregation and analysis module. The intelligent injection assistance module is configured to break down the injection process of biological agents into multiple consecutive steps, send operation guidance data for the current step to the user mobile terminal, and after receiving the completion confirmation instruction for the current step from the user mobile terminal, send operation guidance data for the next sequential step to the user mobile terminal. The compliance control module is configured to send an injection reminder instruction to the user's mobile terminal according to a preset injection cycle; receive the actual injection time and actual injection dose uploaded by the user's mobile terminal; calculate the deviation time between the actual injection time and the preset injection time, and the dose deviation value between the actual injection dose and the preset single dose. Record the number of missed injections, the duration of the deviation, and the dose deviation value to generate a medication log; The adverse reaction monitoring module is configured to receive adverse reaction symptom feedback data and vital sign data uploaded by the user's mobile terminal at preset time nodes. The adverse reaction symptom feedback data includes text description data and injection site image data. Natural language processing algorithms are used to extract symptom keywords from the text description data and match them with a preset adverse reaction symptom database. Image recognition algorithms are used to extract local reaction feature parameters from the injection site image data. Based on the symptom keyword matching results, the local reaction feature parameters, and the abnormality level of the vital sign data, the adverse reaction risk level is determined to be low risk, medium risk, or high risk. The multi-level early warning push module is configured to use the number of missed injections, the duration of deviation, the dose deviation value, and the adverse reaction risk level as conditional variables to execute the early warning judgment logic; when it is determined that the low-level early warning triggering conditions are met, a low-level early warning instruction is sent to the user's mobile terminal. When it is determined that the conditions for triggering a medium-level warning are met, a medium-level warning instruction is issued to the user mobile terminal and the medical staff management terminal. When the conditions for triggering a high-level warning are met, a high-level warning instruction is issued to the medical staff management terminal. The data aggregation and analysis module is configured to aggregate the medication log and the adverse reaction risk level according to a set time period, and extract and store the calculated number of missed injections, injection timeliness rate parameters and frequency of adverse reactions in the database.

2. The home-based biological agent injection compliance and adverse reaction early warning system for inflammatory bowel disease according to claim 1, characterized in that, The multiple step nodes, in the order of operation time, include the drug preparation node, site selection node, skin disinfection node, injection execution node, post-injection pressing node, and waste disposal node. The intelligent injection assistance module is also configured to respond to a node return command issued by the user's mobile terminal and reissue the operation guidance data of the completed step nodes located before the current step node; after the intelligent injection assistance module receives the completion confirmation command for the complete injection process a total of three times within a recorded continuous injection cycle, it grants the user's mobile terminal permission to skip operation commands for each of the step nodes.

3. The home-based biological agent injection compliance and adverse reaction early warning system for inflammatory bowel disease according to claim 1, characterized in that, The execution logic of the compliance control module in sending injection reminder instructions and recording the number of missed injections is as follows: One day and 30 minutes before the scheduled injection time, respectively, a first-round reminder instruction and an upcoming reminder instruction are sent to the user's mobile terminal. If the actual injection time is not received 8 hours after the scheduled injection time, an initial delay reminder instruction is generated and sent. If the actual injection time is not received 12 hours after the scheduled injection time, a reminder instruction to delay again will be generated and sent. If the actual injection time is not received within 24 hours after the scheduled injection time, the current record is marked as a missed injection, the number of missed injections is accumulated, and a remote intervention notification instruction is sent to the medical management terminal.

4. The home-based biological agent injection compliance and adverse reaction early warning system for inflammatory bowel disease according to claim 1, characterized in that, The preset time nodes are, in order, within 30 minutes, 24 hours, and 72 hours after a single actual injection. When the adverse reaction monitoring module determines that the preset time node has been reached and adverse reaction feedback data has not been obtained within the time limit, it triggers a pop-up selection command to the user's mobile terminal. The vital signs data include body temperature data and a set of data items consisting of resting heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate, and blood oxygen saturation. The adverse reaction monitoring module has a built-in rule for classifying abnormal vital signs, which divides the acquired vital sign data into three levels: normal range, mild abnormality (Level 1), moderate abnormality (Level 2), and severe abnormality (Level 3). The adverse reaction monitoring module is configured to, when the received vital sign data includes the body temperature data item but the other data items in the data item set are missing, disable the abnormality judgment and comparison step for the missing data item, and output the adverse reaction risk level only based on the abnormality level of the body temperature data item, the matching result of the symptom keywords, and the local reaction characteristic parameters.

5. The home-based biological agent injection compliance and adverse reaction early warning system for inflammatory bowel disease according to claim 4, characterized in that, The adverse reaction monitoring module determines that the risk level of the adverse reaction is low under the specific condition that the following set of elements are met simultaneously: The local reaction characteristic parameters indicate that the erythema diameter is less than 5 cm and there is no induration feature; The symptom keyword matching results do not contain words corresponding to the systemic symptom category; All of the vital signs data are within the normal range, or the number of data items in the vital signs data that are at the first abnormal level is less than or equal to two and there are no data items that are at the second or third abnormal level.

6. The home-based biological agent injection compliance and adverse reaction early warning system for inflammatory bowel disease according to claim 4, characterized in that, The specific condition for the adverse reaction monitoring module to determine that the adverse reaction risk level is medium risk is that any of the following conditions are met: The local reaction feature parameters indicate that the erythema diameter is in the range of greater than or equal to 5 cm and less than or equal to 10 cm, or the extraction result contains induration features; The symptom keyword matching results matched the mild systemic symptom classification thesaurus. The number of data items in the vital signs data that fall into the second abnormality level is one or two; The number of data items in the vital signs data that fall at the first abnormal level is greater than or equal to three.

7. The home-based biological agent injection compliance and adverse reaction early warning system for inflammatory bowel disease according to claim 4, characterized in that, The specific condition for the adverse reaction monitoring module to determine that the adverse reaction risk level is high is that any of the following conditions are met: The local reaction characteristic parameters indicate that the erythema diameter is greater than 10 cm, or the extraction results include features of blisters, ulceration or necrosis. The symptom keyword matching results matched the severe systemic symptom classification thesaurus. The vital signs data include data items at the third abnormality level; The number of data items in the vital signs data that are at the second abnormal level is greater than or equal to two; The adverse reaction monitoring module is configured with a high-risk direct trigger path: when the extracted symptom keywords match words in the preset emergency danger word library, or when a large area of ​​necrosis feature value is extracted from the local reaction feature parameters, the adverse reaction monitoring module directly outputs the high-risk judgment result and suspends the abnormality level judgment logic thread of the vital signs data.

8. The home-based biological agent injection compliance and adverse reaction early warning system for inflammatory bowel disease according to claim 1, characterized in that, The warning triggering logic set by the multi-level warning push module includes: The low-level warning triggering conditions include any one of the following: the missed injection status is the first single missed injection, the deviation duration is greater than zero and less than or equal to 24 hours of delayed record, the dose deviation value is non-zero and input data containing correction coverage action is obtained, or the adverse reaction risk level is low risk. The triggering conditions for the medium-level warning include any one of the following: the number of missed injections reaches twice within two consecutive judgment periods, the deviation duration is greater than 24 hours and less than the preset latest supplementary injection time threshold, the dose deviation value is obtained twice consecutively as non-zero, or the adverse reaction risk level is medium risk.

9. The home-based biological agent injection compliance and adverse reaction early warning system for inflammatory bowel disease according to claim 8, characterized in that, The warning triggering logic set by the multi-level warning push module also includes: The high-level warning triggering conditions include any one of the following: the duration of a single missed injection is greater than the latest injection time threshold; missed injection status is recorded in three or more consecutive judgment periods; the duration of the deviation obtained in a single instance is greater than the latest injection time threshold; a non-zero dose deviation value is obtained three or more times; or the adverse reaction risk level is high risk. The high-level warning command is configured to, upon being received by the medical staff management terminal, execute a forced pop-up interrupt command on the interface and activate the emergency voice dialing call interface.

10. The home-based biological agent injection compliance and adverse reaction early warning system for inflammatory bowel disease according to claim 1, characterized in that, The data aggregation and analysis module is also configured to: The intervention record data, which includes communication timestamps, interaction method tags, and processing opinion text, is received by the medical care management terminal in response to the intermediate or high-level warning command. The intervention record data is mapped and bound to the medication log and adverse reaction risk level of the data source that triggered the corresponding warning command based on foreign keys. Extract compliance quantitative scoring variables and classification distribution matrix sets of adverse reactions of different risk levels within a set historical time window. Calculate the correlation between the number of missed injections and the frequency of dosage errors and the distribution of adverse reaction risk levels. Update the classification feature set of the preset adverse reaction symptom database by calculating the correlation weights returned from the calculation, and modify the delay threshold tolerance parameters used to determine the triggering conditions for each level of warning accordingly.

Images