Medication-related osteonecrosis of the jaw three-tier risk stratification management system

Abstract

The present application relates to the technical field of medical risk management, in particular to a drug-related jaw necrosis three-level risk stratification management system, aiming to solve the problems of risk assessment fragmentation, stratification standard ambiguity, poor multidisciplinary cooperation, large image assessment error and the like in traditional MRONJ prevention and control. The system comprises a data support layer, a core function layer, an application layer, an emergency response layer and a guarantee layer: the data support layer integrates multi-dimensional clinical data and adapts to basic equipment; the core function layer realizes dynamic risk assessment, three-level differentiated intervention, intelligent image analysis and MDT closed-loop cooperation through fusion algorithm; the application layer adapts to medical staff, patients and management multi-terminal; the emergency response layer handles clinical emergencies; and the guarantee layer ensures system compliance and stability. The system can improve the accuracy of risk assessment, realize individualized diagnosis and treatment, optimize the allocation of medical resources, adapt to the scenes of tertiary and primary hospitals, effectively reduce the incidence of MRONJ, and improve patient compliance and diagnosis and treatment efficiency.

Description

Technical Field

[0001] This invention relates to the field of medical risk management technology, specifically a three-tiered risk management system for drug-related osteonecrosis of the jaw. Background Technology

[0002] Drug-associated osteonecrosis of the jaw (MRONJ) is a serious complication of bisphosphonates used to treat osteoporosis and bone metastases from malignant tumors. Its incidence is particularly high, reaching 50%-80% after invasive procedures such as tooth extraction. Clinical manifestations include jawbone exposure, infection, pain, and even pathological fractures, severely impacting patients' quality of life. Clinical data show that the incidence of MRONJ in cancer patients after intravenous bisphosphonate injection can reach 0.7%-9%, significantly higher than the 0.02%-0.04% in patients receiving oral osteoporosis treatment. Furthermore, 60%-70% of cases are directly related to invasive procedures, while 30%-40% develop spontaneously without any obvious cause.

[0003] Currently, MRONJ prevention and control faces several technical bottlenecks: First, risk assessment is fragmented. Traditional methods rely on single indicators or physician experience, failing to integrate multi-dimensional information such as medication history (including drug type), underlying diseases (including comorbidities), age, imaging characteristics, and laboratory data, leading to inaccurate risk prediction and strong subjectivity. Second, stratified management standards are vague, lacking unified risk grading thresholds and failing to consider the differences in efficacy of different bisphosphonate drugs, the cumulative effect of comorbidities, and age factors. Low-risk patients may receive overtreatment, while intervention for high-risk patients is delayed, hindering personalized treatment. Third, patient compliance is low. Traditional follow-up relies on manual management, lacking accuracy in medication reminders and follow-up appointments, and is not adapted to the operational capabilities of elderly patients, resulting in high rates of missed diagnosis or loss to follow-up. Fourth, multidisciplinary team (MDT) collaboration is ineffective, with untimely communication and unclear division of responsibilities among departments such as oral surgery, endocrinology, and radiology. Expert resources are scarce in primary and remote areas, and consultation response for high-risk patients is delayed. Fifth, imaging assessment has large errors. Traditional CBCT interpretation relies on physician experience and lacks standardized procedures and quantitative analysis, resulting in low detection rates of early microlesions. Furthermore, the poor image quality output from low-configuration equipment in primary hospitals further affects the accuracy of assessment. Sixth, emergency response mechanisms are lacking. No closed-loop response process has been established to address issues such as poor collaboration, imaging errors, insufficient compliance, data security, and a shortage of expert resources, affecting the effectiveness of intervention. Seventh, the technical threshold is high and the promotion cost is high. Primary hospitals lack relevant equipment, professional and technical personnel, and funding, making it difficult for the system to be widely implemented.

[0004] While existing research has proposed some risk assessment tools and stratification concepts, such as the FRAX fracture risk assessment tool and periodontal screening scores, an integrated system has not been formed, resulting in shortcomings such as single indicators, low algorithm reliability, poor functional integration, and insufficient adaptability. For example, research from Peking University School of Stomatology indicates that MRONJ lesions have pathological stratifications such as inflammatory areas and sclerotic areas, but traditional treatments have not incorporated quantitative imaging characteristics into intervention plans; research from Beijing Anzhen Hospital affiliated with Capital Medical University shows that zoledronic acid users have poor healing after tooth extraction, but a risk stratification standard based on medication duration, drug type, and imaging characteristics has not been established.

[0005] Therefore, there is an urgent need to develop an integrated management system that integrates multi-dimensional data, has precise stratification, intelligent collaboration, emergency response, and multi-scenario adaptability to address the shortcomings of existing technologies. Summary of the Invention

[0006] This invention aims to provide a three-tiered risk stratification management system for drug-related osteonecrosis of the jaw (MRONJ), which solves the problems of fragmented risk assessment, vague stratification standards, poor multidisciplinary collaboration, large imaging assessment errors, lack of emergency treatment, insufficient patient-side adaptation, and high promotion costs in traditional risk assessment. It achieves accurate quantification of MRONJ risk, individualized intervention, dynamic monitoring, full-cycle quality control, and adaptability and promotion in multiple scenarios (tertiary hospitals and primary hospitals).

[0007] The technical solution adopted by this invention to solve its technical problem is: a three-level risk stratification management system for drug-related osteonecrosis of the jaw, including a data support layer, a core function layer, an application layer, an emergency response layer, and a protection layer; The data support layer includes a data acquisition module, a data processing module, a weight determination module, and a primary care adaptation module. The data acquisition module integrates basic clinical data, risk assessment data, imaging data, and laboratory data, and supports switching between basic and full data acquisition modes. The data processing module performs data cleaning, standardization, real-time updates, and encrypted storage. The weight determination module determines the weights of indicators through the Delphi method and multivariate logistic regression analysis. The primary care adaptation module supports data import from low-configuration devices and simplifies data acquisition. The core functional layer includes a dynamic risk assessment module, a three-level risk stratification management module, a remote image intelligent analysis module, an intelligent follow-up and health education module, and a multidisciplinary team (MDT) module. The dynamic risk assessment module uses a random forest and logistic regression fusion algorithm. The three-level risk stratification management module formulates differentiated intervention strategies based on evidence-based thresholds combined with drug type, comorbidities, and age factors. The remote image intelligent analysis module combines CNN algorithms with a double-blind review mechanism. The MDT module constructs a closed-loop collaborative process. The application layer includes medical staff, patients, and management terminals, enabling multi-terminal data synchronization and functional adaptation; the emergency response layer establishes a closed-loop processing flow to address issues such as poor multidisciplinary collaboration, image assessment errors, insufficient patient compliance, data security risks, and a lack of expert resources. The safeguards layer covers ethical compliance, quality control, and technical support to ensure the system operates stably and in compliance with regulations.

[0008] Specifically, the clinical basic data in the data acquisition module includes comprehensive data on bisphosphonate use (including drug type and cumulative dosage), systemic underlying diseases (including the number and control level of comorbidities), and age stratification information. Risk assessment-specific data includes FRAX scores and quantitative data from intraoral clinical examinations. Imaging data is compatible with DICOM standards and the output format of low-configuration CBCT equipment at the primary care level, as well as quantitative data on lesion characteristics. Laboratory data includes serum CTX, PINP, inflammatory markers, and coagulation function data. The basic acquisition mode only retains core indicators (medication history, FRAX score, key imaging features, and underlying diseases).

[0009] Specifically, the dynamic risk assessment module's assessment indicator system includes 20 indicators across 6 dimensions: fracture risk, medication risk (including drug type weight), underlying disease risk (including comorbidity superposition coefficient), local oral risk, imaging risk, and age-stratified risk. The risk score ranges from 0 to 100 points and features risk attribution analysis and "risk-benefit balance" medication discontinuation decision support functions (integrating quantitative parameters of tumor condition and osteoporosis severity).

[0010] Specifically, the low-risk threshold for the three-tiered risk management module is: FRAX score <10%, medication duration <3 years or <1 year (oral / intravenous), risk score <30, no severe comorbidities, and age <75 years; medium-risk is: FRAX score 10%-20%, medication duration 3-5 years or 1-2 years (intravenous), risk score 30-60, 1-2 mild comorbidities, and age 75-80 years, or medication duration of a specific bisphosphonate drug (such as zoledronic acid) ≥2 years (oral / intravenous) ≥6 months; high-risk is: FRAX score >20%, medication duration >5 years (oral / intravenous) >2 years, risk score >60, ≥1 severe comorbidity, and age >80 years, or medication duration of a specific bisphosphonate drug (such as zoledronic acid) ≥4 years (oral / intravenous) ≥1 year.

[0011] Specifically, the remote image intelligent analysis module can automatically optimize the image quality of low-configuration devices, quantify and label the extent of bone sclerosis and the size of dead bone, enhance the recognition algorithm for early minor lesions, initiate third-party expert arbitration when there are discrepancies in double-blind review, support historical image comparison analysis, and switch to simplified analysis mode (focusing on core lesion features) in grassroots scenarios.

[0012] Specifically, the MDT module covers oral surgery, endocrinology, radiology, periodontology, pharmacy and infectious diseases, and establishes a regional expert database. High-risk patients can trigger a consultation request within 24 hours, and a joint consultation can be completed and a standardized plan generated within 48 hours. Primary hospitals can apply for cross-regional consultation support through the expert database.

[0013] Specifically, the emergency response layer addresses issues with poor multidisciplinary collaboration by sequentially triggering reminders from department heads, supervision from the medical affairs department, and intervention from the emergency expert team; it initiates third-party arbitration for image assessment errors and activates remote support from higher-level hospitals when no suitable experts are available in the expert database; it employs a tiered intervention approach to address insufficient compliance, including intelligent reminders, manual follow-ups, family assistance, and compliance reward mechanisms; it adapts to different regional medical data privacy regulations and strengthens the encryption level of cross-regional data transmission to address data security risks; and it establishes a tertiary hospital expert resident support mechanism to address the shortage of expert resources.

[0014] Specifically, the intelligent follow-up and health education module supports multi-channel reminders (mobile, SMS, voice), pushes personalized content to patients (for low-risk patients, pushes oral hygiene maintenance videos; for medium- and high-risk patients, pushes early symptom identification text and audio explanations), provides simplified popular science content for elderly patients and patients with low literacy levels, and supports patients to report their condition and obtain preliminary risk assessments through mobile devices and telephone.

[0015] Specifically, the quality control module of the protection layer updates the risk assessment model weights every 6 months, adding weight optimization factors for comorbidities, drug types, and age stratification; the technical support module achieves seamless integration with hospital HIS / LIS / PACS systems and primary care simplified HIS systems, providing modular deployment solutions (basic, standard, and full versions), data transmission uses the AES-256 encryption algorithm, and cross-regional transmission adds an additional desensitization process; technical support includes 24 / 7 remote troubleshooting and online practical training functions for primary care medical staff.

[0016] Specifically, the management system has functions such as real-time monitoring of quality control indicators, evaluation of system algorithm performance, analysis of regional risk trends, and monitoring of promotion costs. It can automatically generate statistical reports such as MRONJ incidence rate, intervention timeliness rate, and grassroots adaptation and usage rate, and supports modular function enable / disable management.

[0017] The beneficial effects of this invention are: Enhancing the accuracy and scientific rigor of risk assessment: The system overcomes the fragmented limitations of traditional risk assessment by integrating multi-dimensional information, including basic clinical data (comprehensive medication information, underlying diseases, and age), risk assessment-specific data, imaging data, and laboratory data. It constructs an assessment system encompassing 6 dimensions and 20 indicators, and uses a random forest and logistic regression fusion algorithm to generate risk scores. Simultaneously, it features risk attribution analysis and "risk-benefit balance" medication discontinuation decision support functions. No longer relying on single indicators or physician experience, it helps medical staff more comprehensively and objectively quantify patient risk, providing a scientific basis for treatment decisions.

[0018] To achieve individualized stratified intervention and optimize the allocation of medical resources, the system, based on evidence-based thresholds and combined with factors such as drug type (e.g., specific bisphosphonates), severity of comorbidities, and age, clearly classifies patients into low, medium, and high risk levels and develops targeted, differentiated intervention strategies. This approach avoids overtreatment for low-risk patients while ensuring timely and intensive intervention for high-risk patients, effectively addressing the imbalance in treatment caused by the ambiguity of traditional stratification standards, and achieving the goal of precise matching and on-demand individualized treatment.

[0019] Breaking down barriers to multidisciplinary collaboration and improving the efficiency of diagnosis and treatment coordination: The system's MDT module constructs a closed-loop collaborative process covering multiple departments such as oral surgery, endocrinology, and radiology, and establishes a regional expert database. Consultation requests can be quickly triggered for high-risk patients, ensuring timely generation of treatment plans. Primary hospitals can also apply for cross-regional consultation support through the expert database, effectively alleviating the problem of uneven distribution of expert resources, reducing treatment delays caused by poor interdepartmental communication and unclear division of responsibilities, and improving the standardization and efficiency of multidisciplinary collaboration.

[0020] Improving the standardization and accuracy of image assessment: The remote intelligent image analysis module enhances the ability to identify early, minute lesions through CNN algorithms. It can automatically optimize image quality from low-configuration equipment at the primary care level, quantify and label lesion characteristics (such as the extent of bone sclerosis and the size of sequestra), and is equipped with a double-blind review and third-party arbitration mechanism. It also supports comparative analysis of historical images. This addresses the problems of traditional image assessment, such as reliance on physician experience, insufficient standardization, and low detection rate of early lesions, and promotes the development of image interpretation towards standardization and precision.

[0021] Enhancing patient compliance and reducing missed diagnoses and loss to follow-up: The intelligent follow-up and health education module supports reminders through multiple channels such as mobile devices, SMS, and voice. It can push personalized content based on the patient's risk level (e.g., oral hygiene guidance for low-risk patients, symptom identification information for medium- and high-risk patients), and provides simplified information for elderly patients and patients with low literacy levels. Simultaneously, through a tiered compliance intervention mechanism (intelligent reminders, manual follow-up, and family assistance, etc.), it improves patient cooperation with medication and follow-up examinations, effectively reducing the probability of missed diagnoses and loss to follow-up, and ensuring continuity throughout the entire treatment cycle.

[0022] A robust emergency response mechanism ensures the stability of diagnosis and treatment: The emergency response layer addresses common clinical issues such as poor multidisciplinary collaboration, imaging assessment errors, insufficient patient compliance, data security risks, and a shortage of expert resources. It establishes a tiered, closed-loop processing procedure (e.g., triggering departmental alerts and medical affairs department supervision in case of poor collaboration; initiating third-party arbitration in case of imaging errors). This mechanism can quickly respond to and resolve unexpected problems during diagnosis and treatment, preventing intervention interruptions or reduced effectiveness due to a lack of emergency response, thus ensuring the stability and reliability of the treatment process.

[0023] Adapting to diverse healthcare needs and promoting grassroots prevention and control: The system supports data import from low-configuration devices and simplifies data collection through a grassroots adaptation module. It offers basic, standard, and full modular deployment solutions, seamlessly integrating with simplified HIS systems in grassroots hospitals. Simultaneously, it includes online practical training for grassroots medical staff and an expert on-site support mechanism, effectively reducing the usage threshold and promotion costs for grassroots hospitals. This solves the problems of high technical barriers and difficulty in grassroots implementation associated with traditional systems, helping grassroots hospitals improve their MRONJ prevention and control capabilities and extending prevention and control work to the grassroots level.

[0024] Ensuring data security and long-term system operation: The security layer uses the AES-256 encryption algorithm to encrypt and store data, and adds additional de-identification processing when transmitting data across regions to comply with the medical data privacy regulations of different regions; the quality control module updates the risk assessment model weights every 6 months to ensure that the algorithm performance is in sync with clinical needs; the technical support module provides 24 / 7 remote troubleshooting to ensure stable and compliant system operation and provide safe and reliable technical support for long-term clinical application.

[0025] To support overall prevention and control management and improve regional prevention and control levels: The management end has functions such as real-time monitoring of quality control indicators, evaluation of system algorithm performance, and analysis of regional risk trends. It can automatically generate MRONJ-related statistical reports to help managers clearly understand the current status of prevention and control in the region (such as the timeliness of intervention and the adaptation and use rate at the grassroots level). This facilitates the optimization of resource allocation, adjustment of prevention and control strategies, and promotes the standardized and scientific management of MRONJ prevention and control work at the regional level, thereby improving the overall level of prevention and control. Attached Figure Description

[0026] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0027] Figure 1 Architecture diagram of the three-level risk stratification management system for drug-related osteonecrosis of the jaw provided by the present invention; Figure 2 Technical roadmap for the three-level risk stratification management system for drug-related osteonecrosis of the jaw provided by this invention; Figure 3 The table of three-level risk stratification intervention strategies for the three-level risk stratification management system for drug-related osteonecrosis of the jaw provided by the present invention. Detailed Implementation

[0028] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0029] like Figures 1-3 As shown, the three-tiered risk management system for drug-related osteonecrosis of the jaw described in this invention includes a data support layer, a core function layer, an application layer, an emergency response layer, and a security layer. Data Support Layer: As the foundation of the system, the data acquisition module integrates basic clinical data (full-dimensional medication information including drug type, underlying diseases including comorbidities, and age stratification), risk assessment data (FRAX scores, quantitative data from intraoral examinations), imaging data (compatible with DICOM standards and low-configuration equipment formats at the grassroots level), and laboratory data (serum CTX, inflammatory markers, etc.), supporting switching between basic and full acquisition modes (basic version is prioritized for grassroots levels). The data processing module uses rule-based filtering and manual review to clean the data, unifying standardized criteria, setting real-time update triggers, and storing data locally and backed up in the cloud, with de-identification and encryption. Desensitization processing is added for cross-regional transmission. The weight determination module determines preliminary weights through three rounds of Delphi method by 15-20 senior experts, and optimizes them using Logistic regression combined with multi-center data (including subgroups of different ages, comorbidities, and drug types) to ensure that the indicator weights are strongly correlated with MRONJ risk. The grassroots adaptation module optimizes the low-configuration CBCT data import interface, simplifies data acquisition fields, and enables integration with the basic-level HIS system.

[0030] Core Functional Layer: This is the system's core engine. The dynamic risk assessment module uses a random forest and logistic regression fusion algorithm to generate a 0-100 risk score and attribution analysis based on 20 indicators across 6 dimensions. It is complemented by a "risk-benefit balance" medication discontinuation decision aid (integrating quantitative parameters of tumor condition and osteoporosis severity). The three-tiered risk management module classifies risks into low, medium, and high based on evidence-based thresholds combined with drug type, comorbidities, and age, and formulates differentiated intervention strategies (such as shortening the intervention period with specific drugs and strengthening monitoring for patients with comorbidities), achieving precise adaptation of medication adjustments, oral health management, and follow-up frequency. The remote image intelligent analysis module... The module enhances the identification of early, minor lesions through CNN algorithms, quantifies and labels lesion characteristics, reduces errors through double-blind review, third-party arbitration, and expert database support, supports historical image comparison, and simplifies analysis modes when switching to grassroots scenarios; the intelligent follow-up and health education module supports reminders through multiple channels such as mobile devices, SMS, and voice, and pushes personalized content (simplified popular science articles adapted for elderly / low-literacy patients), realizing functions such as medication reminders, follow-up appointments, and disease feedback (mobile devices and telephone); the MDT module constructs a closed-loop collaborative process for multiple departments, establishes a regional expert database, triggers consultations for high-risk patients within 24 hours, completes the treatment plan within 48 hours, and grassroots hospitals can apply for cross-regional consultations.

[0031] Application Layer: Enables multi-terminal adaptation. The medical staff terminal supports data entry (basic / full version switching), risk assessment, MDT consultation, image reading, and other functions. It adds risk trend analysis and operation log traceability. The primary care version simplifies the operation interface and adds guided data entry. The patient terminal has multi-channel accurate reminders, personalized education (adapted to different population groups), disease feedback (multiple methods), and appointment booking functions. The management terminal implements permission management, quality control, data analysis, technology promotion, and cost monitoring functions, and supports modular function management.

[0032] Emergency Response Layer: To address issues such as poor multidisciplinary collaboration, imaging assessment errors, insufficient patient compliance, data security risks, and a shortage of expert resources, a tiered response process is established (e.g., activating support from higher-level hospitals when experts are scarce) to ensure timely resolution of problems and avoid intervention delays.

[0033] The safeguard layer covers ethical compliance (electronic informed consent, data privacy protection, and adaptation to different regional regulations), quality control (model calibration every 6 months, with added weights for comorbidities, drug types, and age stratification), and technical support (modular system deployment, 24 / 7 fault response, and online training for primary care physicians) to ensure stable and compliant system operation and reduce promotion costs.

[0034] Example 1: Routine application scenario in tertiary hospitals This embodiment is applied to the oral surgery department of a hospital to verify the effectiveness of the system in routine applications in a tertiary hospital.

[0035] Patient selection: 120 patients who met the diagnostic criteria for osteoporosis or bone metastasis of malignant tumors and were receiving bisphosphonate therapy were included and randomly divided into an observation group (60 cases) and a control group (60 cases). Inclusion criteria: age ≥18 years, normal vision and hearing, able to cooperate with examinations and follow-ups, and signed informed consent; Exclusion criteria: history of epilepsy / mental illness, severe skeletal disease, use of drugs affecting the study within 3 months, drug allergy or contraindications, and participation in other clinical trials.

[0036] The observation group consisted of 28 males and 32 females, aged 45-85 years, with a mean age of (63.5±9.2) years. Among them, 42 patients had osteoporosis, 18 had bone metastases from malignant tumors, 25 had comorbidities (1-3 types), and 15 were using zoledronic acid. The control group consisted of 26 males and 34 females, aged 43-84 years, with a mean age of (62.8±9.5) years. Among them, 40 patients had osteoporosis, 20 had bone metastases from malignant tumors, 23 had comorbidities (1-3 types), and 13 were using zoledronic acid.

[0037] System operation procedure: (1) Data collection: The observation group entered the patient's basic information, bisphosphonate medication data (type, route, dosage, duration), systemic underlying diseases and their control level (including comorbidities), age stratification, FRAX score, intraoral examination data (depth of gingival fissure, range of mucosal swelling, etc.) through the system data collection module (full version), and uploaded CBCT images (DICOM format) and laboratory data (serum CTX, CRP, coagulation function, etc.); the control group collected medical history, intraoral examination and panoramic radiograph data in a conventional manner.

[0038] (2) Risk assessment: After automatic cleaning and standardization of data in the observation group, risk scores and levels were generated by a random forest and logistic regression fusion algorithm, and risk attribution analysis was output. Among them, 28 patients had risk scores <30 (low risk), 22 had scores between 30 and 60 (medium risk), and 10 had scores >60 (high risk); in the control group, the risk level was determined by the physician's experience, with 30 cases of low risk, 22 cases of medium risk, and 8 cases of high risk.

[0039] (3) Stratified intervention: Patients in the low-risk observation group maintained routine medication, had an oral examination and CBCT re-examination once a year, and received oral hygiene guidance via intelligent follow-up every 6 months; patients in the medium-risk group (including those who had been taking zoledronic acid for ≥2 years) had their medication dosage adjusted in conjunction with the endocrinology department, had an oral examination every 3 months and a CBCT re-examination every 6 months, and received intelligent follow-up and risk self-assessment every 3 months; patients in the high-risk group (including those with 2 or more severe comorbidities and those over 80 years of age) immediately discontinued bisphosphonate medication, triggered an MDT application within 24 hours, completed a multidisciplinary consultation involving oral surgery, endocrinology, and radiology within 48 hours, performed jawbone curettage and anti-infection treatment, received a remote CBCT assessment once a month, and received intelligent follow-up every week and manual follow-up every 2 weeks. The control group received routine intervention, with medication adjusted based on physician experience, and received regular manual follow-up.

[0040] (4) Emergency response: In the observation group, the Department of Endocrinology did not respond in a timely manner during the MDT consultation for one high-risk patient. The system automatically reminded the department head and completed the response within 12 hours. There were disagreements in the CBCT imaging assessment of two patients. Third-party expert arbitration was initiated and a final opinion was issued within 48 hours. Three patients showed insufficient compliance. The system pushed warning information and family members helped remind them. Medical staff followed up by phone and cooperated with the compliance reward mechanism to improve compliance.

[0041] Application effects: The incidence of MRONJ in the observation group was 2.00% (1 / 60), significantly lower than that in the control group (16.67%, 10 / 60); the medication compliance rate in the observation group was 95.00% (57 / 60), and the follow-up completion rate was 93.33% (56 / 60), both higher than those in the control group (75.00%, 45 / 60 and 70.00%, 42 / 60); the average response time for MDT consultations for high-risk patients was 32 hours, significantly shorter than that in the control group (72 hours); the average patient satisfaction score was 90.2, higher than that in the control group (76.2).

[0042] Example 2: Multi-center promotion and application scenario This embodiment was applied to three tertiary hospitals to verify the applicability of the system for multi-center promotion.

[0043] Patient selection: 80 patients were included from each hospital, for a total of 240 patients, all of whom met the inclusion and exclusion criteria in Example 1.

[0044] Among them, Hospital A (the Third Affiliated Hospital of Air Force Medical University) had 38 males and 42 females, aged 44-86 years, with an average age of (64.1±8.5) years; Hospital B (Peking University School of Stomatology) had 36 males and 44 females, aged 45-85 years, with a mean age of (63.5±8.8) years. Hospital C (West China Hospital of Stomatology, Sichuan University) included 39 males and 41 females, aged 42-87 years, with a mean age of (65.2±9.3) years.

[0045] All patients were managed using the system of this invention, including 102 patients with comorbidities and 115 patients using specific bisphosphonate drugs (zoledronic acid, alendronate).

[0046] System operation procedure: (1) System deployment: Deploy the system (standard version) in 3 hospitals, complete the connection with the HIS / LIS / PACS system of each hospital, and provide training to medical staff on system operation, risk assessment standards and MDT process. After passing the assessment, they can start working.

[0047] (2) Data collection and sharing: Each hospital collects multi-dimensional patient data through the system. After de-identification and desensitization, multi-center data sharing is achieved. The weight determination module optimizes the indicator weights based on data from 240 patients (including different drug types, comorbidities, and age subgroups).

[0048] (3) Risk assessment and intervention: The system generates risk levels according to unified standards, and each hospital implements interventions according to a stratified strategy. Hospital A has 42 low-risk cases, 28 medium-risk cases, and 10 high-risk cases; Hospital B has 40 low-risk cases, 30 medium-risk cases, and 10 high-risk cases; Hospital C has 41 low-risk cases, 29 medium-risk cases, and 10 high-risk cases. All high-risk patients initiate cross-hospital MDT consultations through the system (e.g., patients from Hospital A invite radiology experts from Hospital B to participate in arbitration).

[0049] (4) Quality control: The management end monitors the incidence of MRONJ, intervention timeliness rate, follow-up completion rate and other indicators of the three hospitals in real time. The system automatically updates the risk assessment model weights every 6 months and verifies the algorithm effectiveness based on multi-center and multi-subgroup data.

[0050] Application Results: The overall incidence of MRONJ in the three hospitals was 3.33% (8 / 240), with Hospital A at 2.50% (2 / 80), Hospital B at 3.75% (3 / 80), and Hospital C at 3.75% (3 / 80). There was no significant difference in incidence, indicating that the system is adaptable to different hospital scenarios. The average response time for multi-center MDT consultations was 28 hours, the accuracy rate of image assessment was 92.50%, medication adherence was 94.17%, and the average patient satisfaction score was 89.5. The system's data entry error rate was 0.6%, and the operational failure rate was 0.4%, demonstrating good stability and applicability.

[0051] Example 3: Application Scenarios Adapted for Primary Hospitals This embodiment was applied to two primary hospitals to verify the system's adaptability in primary healthcare scenarios.

[0052] Patient selection: 60 patients were enrolled from each hospital, totaling 120 patients, meeting the inclusion and exclusion criteria of Example 1. Among them, 90% (108 cases) had osteoporosis, 10% (12 cases) had bone metastases from malignant tumors, and 45 patients had comorbidities. The patients' ages ranged from 46 to 88 years, with a mean age of (66.5 ± 10.5) years. Hospital A had 27 males and 33 females, while Hospital B had 25 males and 35 females. All patients used low-configuration CBCT equipment at the primary care level, and 60% of the patients lacked smartphone operation skills.

[0053] System operation procedure: (1) System adaptation: Deploy the basic version of the system, enable the basic data acquisition mode, optimize the imaging module to support data import and preprocessing of low-configuration CBCT equipment; simplify the operation interface and add guided data entry function; connect with the basic hospital's simplified HIS system to achieve data synchronization.

[0054] (2) Remote support: By accessing the regional expert database, CBCT images of high-risk patients in primary hospitals can be uploaded to higher-level hospitals through the system and remotely reviewed by senior physicians; MDT consultations adopt a simplified online process and apply for cross-regional consultation support through the expert database.

[0055] (3) Patient management: Enable multi-channel reminder function (mainly SMS and voice), push simplified popular science content (text and voice explanation); simplify the follow-up process, adopt SMS and telephone dual-channel reminders to reduce the difficulty of operation for patients, and start a family assistance + reward mechanism for patients with poor compliance.

[0056] (4) Personnel training: Special training will be conducted for grassroots medical staff, focusing on risk assessment standards, basic interpretation of images and emergency response procedures. The training will combine online video and offline hands-on practice to ensure that they master the core functions of the system.

[0057] Application Results: The incidence of MRONJ in two primary hospitals was 4.17% (5 / 120), with Hospital A at 3.33% (2 / 60) and Hospital B at 5.00% (3 / 60). Although higher than that of tertiary hospitals, it was significantly lower than the incidence of traditional methods in primary hospitals (approximately 20%). Medication adherence was 91.67% (110 / 120), follow-up completion rate was 89.17% (107 / 120), and average patient satisfaction was 87.6. The average response time for remote image review was 40 hours, and the average response time for MDT consultation was 36 hours. 90.0% of medical staff were satisfied with the system operation, believing that the system was adapted to the workflow of primary hospitals, reduced the difficulty of diagnosis and treatment, and that modular deployment reduced hospital investment costs.

[0058] Comparison Example Comparison with Example 1: Traditional routine assessment process Sixty patients admitted to the Department of Oral and Maxillofacial Surgery of a hospital were selected. A traditional, routine assessment process was used, including medical history taking, intraoral examination, and panoramic radiography. The risk level was determined by the physician's experience, and medication was adjusted accordingly. Results showed that the incidence of MRONJ was 18.33% (11 / 60), medication adherence was 73.33% (44 / 60), and the follow-up completion rate was 68.33% (41 / 60). The average consultation response time for high-risk patients was 75 hours, and the average patient satisfaction score was 75.6. All indicators were inferior to the system of this invention (Observation Group of Example 1), indicating significant deficiencies in the traditional method.

[0059] Compare with Example 2: Single Algorithm Risk Assessment System Sixty patients admitted to the Department of Oral and Maxillofacial Surgery of a hospital were selected. A risk assessment system based solely on the Logistic regression algorithm was used, without intelligent image analysis, MDT closed-loop management, or emergency response functions. The results showed that the incidence of MRONJ was 8.33% (5 / 60), the accuracy of risk assessment was 76.67%, the error rate of image assessment was 15.00%, the average consultation response time for high-risk patients was 52 hours, and the medication adherence rate was 82.50% (49 / 60). All indicators were lower than those of the system of this invention, indicating that a single algorithm and the lack of core functional modules lead to a decrease in system performance.

[0060] Compare with Example 3: A hierarchical system without an emergency response mechanism Sixty patients admitted to the Department of Oral and Maxillofacial Surgery of a hospital were selected and a system including data collection, risk assessment, and stratified intervention functions but lacking an emergency response mechanism was used. The results showed that the incidence of MRONJ was 7.50% (4 / 60), but the incidence of poor multidisciplinary collaboration was 13.33% (8 / 60), the rate of unresolved discrepancies in imaging assessment was 8.33% (5 / 60), and three cases of intervention failure were due to insufficient patient compliance. This indicates that the lack of an emergency response mechanism affects the stability of the system and the intervention effect.

[0061] The drug-related osteonecrosis of the jaw three-tier risk stratification management system of this invention can realize the functions of each module through computer software programming. Data storage adopts a medical-grade database, and the image analysis algorithm is developed based on the TensorFlow framework of Python. The system interface is jointly developed using HTML5, CSS3, and JavaScript, supporting PC and mobile terminal adaptation, and providing basic, standard, and full version modular deployment solutions. During implementation, the corresponding version should be selected according to the hospital level, the system deployment should be completed and integrated with the hospital's existing system, targeted training should be provided to medical staff (focusing on training primary care medical staff to simplify operating procedures), multi-center, multi-subgroup data should be collected to optimize model weights, and then gradually promoted and applied. During system operation, quality calibration should be performed every 6 months to update the risk assessment model and stratification criteria, and functional modules should be optimized based on user feedback to ensure that the system continuously adapts to the needs of different clinical scenarios.

[0062] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of protection claimed by the present invention. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A three-tiered risk stratification management system for drug-related osteonecrosis of the jaw, characterized in that, It includes a data support layer, a core function layer, an application layer, an emergency response layer, and a security layer; The data support layer includes a data acquisition module, a data processing module, a weight determination module, and a primary care adaptation module. The data acquisition module integrates basic clinical data, risk assessment data, imaging data, and laboratory data, and supports switching between basic and full data acquisition modes. The data processing module performs data cleaning, standardization, real-time updates, and encrypted storage. The weight determination module determines the weights of indicators through the Delphi method and multivariate logistic regression analysis. The primary care adaptation module supports data import from low-configuration devices and simplifies data acquisition. The core functional layer includes a dynamic risk assessment module, a three-level risk stratification management module, a remote image intelligent analysis module, an intelligent follow-up and health education module, and a multidisciplinary collaboration module. The dynamic risk assessment module uses a random forest and logistic regression fusion algorithm. The three-level risk stratification management module formulates differentiated intervention strategies based on evidence-based thresholds combined with drug type, comorbidities, and age factors. The remote image intelligent analysis module combines CNN algorithm and double-blind review mechanism. The multidisciplinary collaboration module constructs a closed-loop collaboration process. The application layer includes medical staff, patients, and management terminals, enabling multi-terminal data synchronization and functional adaptation; the emergency response layer establishes a closed-loop processing flow to address issues such as poor multidisciplinary collaboration, image assessment errors, insufficient patient compliance, data security risks, and a lack of expert resources. The safeguards layer covers ethical compliance, quality control, and technical support to ensure the system operates stably and in compliance with regulations.

2. The three-tiered risk stratification management system for drug-related osteonecrosis of the jaw according to claim 1, characterized in that: The data acquisition module includes clinical basic data such as bisphosphonate medication data, systemic underlying diseases, and age stratification information; risk assessment data such as FRAX score and quantitative data of intraoral clinical examination; imaging data compatible with DICOM standard and low-configuration CBCT equipment output formats at the primary care level; and quantitative data of lesion characteristics. Laboratory data includes serum CTX, PINP, inflammatory markers, and coagulation function data. The basic acquisition mode only retains core indicators.

3. The three-tiered risk stratification management system for drug-related osteonecrosis of the jaw according to claim 1, characterized in that: The dynamic risk assessment module's assessment indicator system includes 20 indicators across 6 dimensions: fracture risk, medication risk, underlying disease risk, local oral risk, imaging risk, and age-stratified risk. The risk score ranges from 0 to 100 points. It has functions such as risk attribution analysis, risk-benefit balance, and medication discontinuation decision support, and integrates quantitative parameters of tumor condition and osteoporosis severity.

4. The three-tiered risk stratification management system for drug-related osteonecrosis of the jaw according to claim 1, characterized in that: The low-risk threshold for the Level 3 risk stratification management module is: FRAX score <10%, medication duration <3 years or <1 year (oral / intravenous), risk score <30, no severe comorbidities, and age <75 years. Medium risk is defined as a FRAX score of 10%-20%, duration of medication (oral) 3-5 years / intravenous 1-2 years, risk score of 30-60, 1-2 mild comorbidities, age 75-80 years, or duration of medication (oral) ≥2 years / intravenous ≥6 months for specific bisphosphonates. High risk is defined as a FRAX score >20%, duration of medication >5 years or >2 years (oral / intravenous), risk score >60, ≥1 severe comorbidity, age >80 years, or duration of medication for a specific bisphosphonate drug ≥4 years (oral / intravenous ≥1 year).

5. The three-tiered risk stratification management system for drug-related osteonecrosis of the jaw according to claim 1, characterized in that: The remote image intelligent analysis module can automatically optimize the image quality of low-configuration devices, quantify and label the extent of bone sclerosis and the size of dead bone, enhance the recognition algorithm for early minor lesions, initiate third-party expert arbitration when there are discrepancies in double-blind review, support historical image comparison analysis, and switch to simplified analysis mode in grassroots scenarios.

6. The three-tiered risk stratification management system for drug-related osteonecrosis of the jaw according to claim 1, characterized in that: The multidisciplinary collaboration module covers oral surgery, endocrinology, radiology, periodontology, pharmacy and infectious diseases, and establishes a regional expert database. High-risk patients can trigger a consultation request within 24 hours, and a joint consultation can be completed and a standardized plan generated within 48 hours. Primary hospitals can apply for cross-regional consultation support through the expert database.

7. The three-tiered risk stratification management system for drug-related osteonecrosis of the jaw according to claim 1, characterized in that: The emergency response layer addresses issues with poor multidisciplinary collaboration by sequentially triggering reminders from department heads, supervision from the medical affairs department, and intervention from the emergency expert team. For image assessment errors, third-party arbitration is initiated; if no suitable expert is available from the expert database, remote support from a higher-level hospital is activated. For insufficient compliance, a tiered intervention approach is adopted, including intelligent reminders, manual follow-ups, family assistance, and compliance reward mechanisms. Regarding data security risks, the layer adapts to different regional medical data privacy regulations and strengthens the encryption level of cross-regional data transmission. Finally, to address the shortage of expert resources, a tertiary hospital expert resident support mechanism is established.

8. The three-tiered risk stratification management system for drug-related osteonecrosis of the jaw according to claim 1, characterized in that: The intelligent follow-up and health education module supports multi-channel reminders, pushes personalized content to patients, provides simplified popular science content for elderly patients and patients with low literacy levels, and supports patients to report their condition and obtain preliminary risk assessments through mobile devices and telephones.

9. The three-tiered risk stratification management system for drug-related osteonecrosis of the jaw according to claim 1, characterized in that: The quality control module of the protection layer updates the risk assessment model weights every 6 months, adding weight optimization factors for comorbidities, drug types, and age stratification; The technical support module achieves seamless integration with hospital HIS / LIS / PACS systems and basic-level simplified HIS systems, providing modular deployment solutions. Data transmission adopts the AES-256 encryption algorithm, and cross-regional transmission adds an additional desensitization process. Technical support includes 24 / 7 remote troubleshooting and online practical training for primary healthcare workers.

10. The three-tiered risk stratification management system for drug-related osteonecrosis of the jaw according to claim 1, characterized in that: The management system features real-time monitoring of quality control indicators, evaluation of system algorithm performance, analysis of regional risk trends, and monitoring of promotion costs. It can automatically generate statistical reports on MRONJ incidence rate, intervention timeliness rate, and grassroots adaptation and usage rate, and supports modular function enable / disable management.

Images