A chronic disease management and clinical interaction record platform for the elderly
By using a chronic disease management platform for the elderly, multi-dimensional health data analysis is conducted using association and causal rules to generate structured clinical summaries and perform feedback optimization. This solves the problem that existing platforms cannot perform multi-dimensional association analysis and causal inference, thereby improving medication safety and treatment efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- THE FIRST PEOPLES HOSPITAL OF NANTONG
- Filing Date
- 2026-03-10
- Publication Date
- 2026-06-12
Smart Images
Figure FT_1
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent management technology, and more specifically, to a chronic disease management and clinical interaction recording platform for the elderly population. Background Technology
[0002] With the accelerating aging of society, the co-occurrence of chronic diseases and multiple medications are becoming increasingly common among the elderly, leading to multidimensional data, complex interactions, and hidden risks in their health management. Traditional health management methods and existing technology platforms are clearly insufficient in addressing this challenge.
[0003] Currently, common methods for chronic disease management mainly include physiological parameter monitoring based on smart wearable devices, medication reminder applications, and remote consultation platforms. Existing technology platforms typically focus on the aggregation and visualization of multi-source health data and can set thresholds for single parameters to trigger alerts when limits are exceeded. Some systems integrate electronic medical record access or doctor-patient messaging functions.
[0004] However, these existing technologies have significant drawbacks: 1. The core logic is mostly based on single-dimensional alarms with fixed thresholds, which cannot perform correlation analysis and causal inference on complex patterns that span multiple types of data such as physiological monitoring, medication records, and subjective symptoms, and cannot reveal the deep medical logic behind abnormal events.
[0005] 2. Its rules and knowledge base are usually statically preset and cannot obtain feedback from continuous clinical use to optimize itself. As a result, the system cannot adapt to individual differences and updates in medical knowledge, and has the rigidity problem of "built once and never changed".
[0006] Therefore, we designed a chronic disease management and clinical interaction recording platform for the elderly population. Summary of the Invention
[0007] The purpose of this invention is to provide a chronic disease management and clinical interaction recording platform for the elderly population, so as to solve the problems mentioned in the background art.
[0008] To achieve the above objectives, the present invention aims to provide a chronic disease management and clinical interaction recording platform for the elderly population, comprising: A data integration and processing unit is used to access and process raw health data from multiple heterogeneous data sources, and to preprocess the raw health data to generate a fused data sequence. The knowledge base and reasoning unit are used to analyze and reason about the input fused data sequence and output a structured reasoning result containing abnormal patterns and inferred medical correlations. A record summary generation unit is used to receive the structured reasoning results and convert the structured reasoning results into natural language clinical summary text based on a predefined clinical summary template. The clinical interaction optimization unit provides a doctor-side review interface for displaying the clinical summary text and receiving doctor review interaction operations, generating clinical verification feedback and transmitting it to the knowledge base and reasoning unit.
[0009] As a further improvement to this technical solution, the original health data includes physiological monitoring data, medication record data, and status data actively input by the elderly patients.
[0010] As a further improvement to this technical solution, the knowledge base and reasoning unit include a knowledge base module, a reasoning engine module, and a rule optimization module; The knowledge base module is used to store a knowledge graph constructed based on medical knowledge and reasoning rules defined based on the logic of the knowledge graph. The reasoning rules include at least association rules and causal rules. The inference engine module is used to call the inference rules to analyze the fused data sequence, perform correlation analysis and causal inference, and output structured inference results; the structured inference results include a first structured inference sub-result and a second structured inference sub-result; The rule optimization module is used to fine-tune the parameters of the historical inference rules that generate the feedback based on the clinical validation feedback from the clinical interaction optimization unit. The parameters include trigger thresholds, association weights, or confidence calculation coefficients.
[0011] As a further improvement to this technical solution, the knowledge graph encodes medical entities, relationships, and pathophysiological logic related to chronic diseases in the elderly using a graph structure. The association rules and causal rules are defined and invoked based on the graph structure logic of the knowledge graph.
[0012] As a further improvement to this technical solution, the inference engine module includes an association analysis submodule, which is configured to perform the following steps to implement association rules: S11. When a specific abnormal pattern from physiological monitoring data, a key drug change event from medication record data, and a relevant status indication from patient input status data are identified simultaneously, correlation analysis is triggered. S12. Based on the predefined medical logic paths in the knowledge graph, the specific abnormal patterns, key drug change events and related status indicators are associated to generate clinical hypotheses describing their potential common pathophysiological basis. S13. Assess the overall risk level of the clinical hypothesis based on the severity and frequency of the specific abnormal pattern and the persistence of related state indicators; S14. Output the first structured reasoning sub-result containing the clinical hypothesis and the comprehensive risk level.
[0013] As a further improvement to this technical solution, the inference engine module includes a causal inference submodule, which is configured to perform the following steps to implement causal rules: S21. When a new drug event is identified in the medication record data, the time of occurrence of the event shall be used as the starting time point. S22. Starting from the aforementioned starting time point, within a preset observation time window, if physiological monitoring data that conforms to the abnormal trend expected by the known abnormal effect causal chain model of the drug in the knowledge graph and symptom reports associated with the patient's input status data are sequentially identified, causal inference is triggered. S23. Call the known abnormal action causal chain model corresponding to the newly added drug in the knowledge graph, verify it and generate a matching score; S24. Proactively query patient data from the same period to identify whether there are competing events that could replace the explanation of the abnormal trends or symptom reports; S25. Based on the matching degree and the exclusion of the competing events, calculate the confidence level of the causal relationship between the new drug event and the subsequent anomaly; S26. Output a second structured inference sub-result containing a description of the causal relationship and the confidence level.
[0014] As a further improvement to this technical solution, in step S23, the matching score is generated as follows: Determine whether the timing of the abnormal trend and the occurrence of symptom reports matches the expectations of the causal chain model; Determine whether the specific clinical manifestations conform to the typical characteristics described by the causal chain model; Based on the agreement between the two judgments, a time series and clinical feature matching score is generated.
[0015] As a further improvement to this technical solution, the clinical summary text includes at least a list of abnormal events arranged in chronological order, a description of the clinical situation based on inferred correlation, and itemized questions to be verified.
[0016] As a further improvement to this technical solution, the doctor-side review interface includes a differentiated information display area and an interactive processing area; The differentiated information display area is used to visualize and associate the inferred medical associations with the original health data; the inferred medical associations include clinical hypotheses generated by the association rules or causal relationships inferred by the causal rules; The interactive processing area provides interactive components for checking and confirming or modifying issues to be verified, as well as functions for confirming or editing clinical summary text.
[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This chronic disease management and clinical interaction recording platform for the elderly achieves proactive and in-depth analysis of clinical conditions through the synergistic effect of association rules and causal rules: association rules proactively discover complex comorbid associations across physiological indicators, medications, and symptoms, effectively avoiding missed diagnoses due to fragmented information; causal rules rigorously verify the causal relationship between drugs and abnormal events using time series and pharmacological models, enabling early and accurate warnings of drug side effects, thereby improving medication safety and risk management capabilities.
[0018] 2. This chronic disease management and clinical interaction record platform for the elderly transforms reasoning results into structured clinical summaries that can be directly used for decision-making, greatly reducing doctors' paperwork burden and improving diagnostic and treatment efficiency. At the same time, it transforms each doctor's review interaction into structured feedback, driving continuous optimization of internal rules and enabling the system to increasingly align with real clinical scenarios and individual patient characteristics. Attached Figure Description
[0019] Figure 1 This is a flowchart illustrating the overall process of the present invention. Detailed Implementation
[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0021] Example: Please refer to Figure 1 As shown, a chronic disease management and clinical interaction recording platform for the elderly is provided, including a data integration and processing unit, a knowledge base and reasoning unit, a record summary generation unit, and a clinical interaction optimization unit; The data integration and processing unit is used to access and process raw health data from multiple heterogeneous data sources, and to preprocess the raw health data to generate a fused data sequence. Raw health data includes physiological monitoring data, medication record data, and status data actively entered by elderly patients; Preprocessing includes cleaning and time-series alignment of health data of elderly users; the cleaning process specifically includes filtering out invalid measurements by applying rules based on the behavioral characteristics of elderly users, including identifying false abnormal values of physiological parameters caused by improper operation; the time-series alignment process unifies data from different sources onto the same time axis to form a standardized data sequence with timestamps. Identifying false abnormal values of physiological parameters caused by improper operation includes: For blood pressure data, analyze the pulse wave signal characteristics within a single measurement cycle. If abnormal characteristics are found (such as drastic changes in waveform amplitude or severe unevenness in cycle), the measurement value is marked as a suspected false abnormality caused by improper operation. For blood glucose data, considering the relationship between blood collection time and meal records, if the blood glucose value is logically significantly inconsistent with the usual postprandial change pattern of the same patient, and the device does not report an error code, then the value is marked as an outlier that should be considered with caution; all marked false outliers are either excluded in subsequent reasoning or given a lower confidence weight.
[0022] The knowledge base and reasoning unit are used to analyze and reason about the input fused data sequence, and output structured reasoning results containing abnormal patterns and inferred medical correlations; the structured reasoning results include a first structured reasoning sub-result and a second structured reasoning sub-result. The knowledge base and reasoning unit includes a knowledge base module, a reasoning engine module, and a rule optimization module; The knowledge base module is used to store a knowledge graph built based on medical knowledge and reasoning rules defined based on the logic of the knowledge graph. The reasoning rules include at least association rules and causal rules. The knowledge graph encodes medical entities, relationships, and pathophysiological logic related to chronic diseases in the elderly using a graph structure. Both association rules and causal rules are defined and invoked based on the graph structure logic of the knowledge graph.
[0023] The specific construction method is as follows: (1) Structured extraction: Using natural language processing technology, medical entities such as “disease”, “symptom”, “drug”, “physiological parameters”, and “test indicators”, as well as semantic relationships such as “cause”, “treatment”, “contraindication”, “monitoring”, and “aggravation” are automatically extracted from clinical guidelines, drug instructions and medical literature to form an initial map skeleton; (2) Expert annotation: Geriatric medicine and clinical pharmacy experts review, revise and enhance the initial map based on the physiological and pathological characteristics of the elderly (such as decreased liver and kidney function, multiple medications, and frailty). Experts manually annotate or strengthen special logical connections and causal chains that are particularly important to the elderly population (e.g., the strong association between "decreased creatinine clearance" and "risk of digoxin poisoning") to ensure the clinical rationality and relevance of the model.
[0024] Furthermore, the association rules are configured to contain one or more predefined medical logic paths, each representing an inference chain from a specific data pattern to a potential clinical conclusion. This addresses the challenge of identifying the common underlying cause when elderly patients present with multiple symptoms and abnormalities. Specifically, the medical logic paths are encoded as production rules in the form of "IF-THEN" and stored in a knowledge base. For example, rule R1: IF: Simultaneous presence of nocturnal hypoglycemia event clusters, use of sulfonylureas, and recent weight loss trend; THEN: Hypothesis – Potential renal insufficiency or inadequate nutritional intake; Risk level = f(hypoglycemia frequency, hypoglycemia severity, weight loss slope). Triggering conditions are set to require the simultaneous occurrence of multiple types of evidence to reduce false positives.
[0025] This allows for the proactive discovery of hidden correlations across data dimensions, integrating scattered anomalies into hypotheses pointing to clear clinical problems, and assisting doctors in differential diagnosis.
[0026] Furthermore, causal rules are used to achieve early, automated monitoring and attribution of adverse drug reactions, especially in elderly populations using multiple drugs in combination. Specifically, they exist in the form of production rules, but their core lies in their inclusion of a complete causal verification and evaluation process, as follows: Model storage: "Known anomalous action causal chain models" are stored in the knowledge graph in the form of frames. For example, for nonsteroidal anti-inflammatory drugs (NSAIDs), their models are stored as follows: [Drug: NSAID] - (can cause) -> [Pathophysiology: water and sodium retention] - (clinical manifestations) -> [Symptoms: elevated blood pressure] (typical incubation period: 3-7 days), [Symptoms: edema] (typical incubation period: 5-14 days); Match score: When a rule is triggered, the system compares the actual observed time series (e.g., blood pressure begins to rise on day 4 after medication, edema is reported on day 10) and clinical manifestations with the model's predictions. The score can be calculated using a preset scoring table or function. A score of 1.0 is awarded for a perfect time sequence; a score of 0.8 is awarded for deviations within a preset tolerance window; and a score of 1.0 is awarded for a perfect symptom match.
[0027] Confidence Calculation: Confidence of Final Causation Relationship It is a function of the matching degree and the exclusion of competing reasons. Optional calculation formulas are:
[0028] In the formula, The weighting coefficients are for competitive reasons. The number of legitimate competitive reasons identified.
[0029] In summary, this rule provides an objective and repeatable method for monitoring adverse drug events, which can output quantitative correlation strength and help clinical decision-making.
[0030] The inference engine module is used to call inference rules to analyze the fused data sequence, perform correlation analysis and causal inference, and output structured inference results; The inference engine module includes an association analysis submodule, which is configured to perform the following steps to implement association rules: S11. When a specific abnormal pattern from physiological monitoring data, a key drug change event from medication record data, and a relevant status indication from patient input status data are identified simultaneously, correlation analysis is triggered. S12. Based on the predefined medical logic paths in the knowledge graph, specific abnormal patterns, key drug change events and related status indicators are associated to generate clinical hypotheses describing their potential common pathophysiological basis. S13. Assess the overall risk level of clinical hypotheses based on the severity and frequency of specific abnormal patterns and the persistence of related state indicators. S14. Output the first structured reasoning sub-result, which includes clinical assumptions and comprehensive risk levels; The inference engine module includes a causal inference submodule, which is configured to perform the following steps to implement causal rules: S21. When a new drug event is identified in the medication record data, the time of occurrence of the event shall be used as the starting time point. S22. Starting from the initial time point, within a preset observation time window, if physiological monitoring data that conforms to the abnormal trend expected by the causal chain model of the known abnormal effects of the drug in the knowledge graph and the symptom reports associated with the patient's input status data are sequentially identified, causal inference is triggered. S23. Call the known abnormal action causal chain model corresponding to the newly added drug in the knowledge graph, verify it and generate a matching score; The matching score is generated as follows: Determine whether the timing of abnormal trends and symptom reports aligns with the expectations of a causal chain model; Determine whether the specific clinical manifestations conform to the typical characteristics described by the causal chain model; Based on the agreement between the two judgments, a score is generated to show the matching degree between the time series and clinical features; Among them, the known abnormal action causal chain model refers to a causal mechanism model stored in the knowledge graph in the form of a frame for a specific drug or drug category, describing the causal mechanism that may lead to abnormal events. Each model contains at least the following structured information: (1) the target drug; (2) the mediated pathophysiological intermediate state (such as water and sodium retention, electrolyte disturbance); (3) the possible abnormal physiological trends and clinical symptoms; (4) the typical temporal relationship and latency range of each clinical event; and (5) the typical characteristics of clinical manifestations. This model is used to perform temporal verification and feature matching of suspected drug-related abnormal events, and to calculate the confidence of causal association through quantitative scoring. Its knowledge sources mainly include drug instructions, drug epidemiology research results and pharmacovigilance databases, and have been reviewed and calibrated by clinical pharmacy experts.
[0031] S24. Proactively query patient data from the same period to identify competing events that could substitute for explanations of abnormal trends or symptom reports; S25. Based on the matching degree and the exclusion of competing events, calculate the confidence level of the causal relationship between the new drug event and subsequent anomalies; S26. Output the second structured inference sub-result, which includes a description of the causal relationship and the confidence level.
[0032] The rule optimization module is used to fine-tune the parameters of the historical inference rules corresponding to the feedback generated by the clinical validation feedback from the clinical interaction optimization unit. The parameters include trigger thresholds, association weights, or confidence calculation coefficients.
[0033] The record summary generation unit receives structured reasoning results and, based on a predefined clinical summary template, converts the structured reasoning results into natural language clinical summary text and automatically generates structured clinical record summaries. The clinical summary text should include at least a chronological list of abnormal events, a description of the clinical situation based on inferred correlations, and itemized questions to be verified; The predefined clinical summary template uses a mechanism that maps machine-readable inference result elements to natural language fragments and fills them in. In the generated summary text, the content inferred by the system is identified using specific wording or format to distinguish it from the description of objective facts.
[0034] The clinical interaction optimization unit provides a doctor-side review interface to display clinical summary text and receive doctors' review interaction operations, and generates clinical validation feedback which is transmitted to the knowledge base and reasoning unit. The data structure of the clinical validation feedback includes summary ID, patient ID, original data conclusion, doctor operation, revised conclusion, and operation timestamp.
[0035] The doctor's review interface includes a differentiated information display area and an interactive processing area; The differentiated information display area is used to visualize the association between the inferred medical associations and the original health data; the inferred medical associations include clinical hypotheses generated by association rules or causal relationships inferred by causal rules; the interactive processing area provides interactive components for checking, confirming or modifying issues to be verified, as well as functions for confirming or editing clinical summary text.
[0036] Finally, the rule optimization module is used to fine-tune the parameters of the historical inference rules corresponding to the feedback generated by the clinical validation feedback from the clinical interaction optimization unit. The parameters include trigger thresholds, association weights, or confidence calculation coefficients.
[0037] This rule optimization module addresses the issues of static and rigid predefined rules, their inability to adapt to individual differences and updates in medical knowledge. The module is implemented based on an incremental learning mechanism using clinical feedback. When the "Clinical Interaction Optimization Unit" sends back clinical validation feedback (such as a doctor modifying a system-generated hypothesis), the rule optimization module is activated. Case Comparison: Extracting the Rules Used to Generate the Original Conclusion Input data and system output The doctor's revised conclusion Compare them.
[0038] Parameter tuning: If Confirmed or reinforced For a particular path (e.g., a doctor confirms the hypothesis of "renal insufficiency"), the rules are fine-tuned using gradient descent. Weights or trigger thresholds for relevant conditions (such as "weight loss trend") are assigned to make it more sensitive to similar patterns in the future.
[0039] Rule discovery: If If a new and reasonable association is introduced (such as a doctor adding "dehydration needs to be checked"), and this pattern frequently appears in historical cases of difference, the module can call the association mining algorithm to suggest generating a new candidate rule, which will be included in the knowledge base after expert review.
[0040] This enables the system to "learn from practice," continuously optimize its performance, and become increasingly aligned with the real clinical scenarios of the target user group, achieving iterative improvements in personalization and precision.
[0041] Example 1: Analysis and optimization of atypical hypoglycemic events based on multimodal association rules. This example demonstrates how the system integrates scattered abnormal signals through association rules to indicate potential common pathophysiological problems.
[0042] Step 11: The data integration and processing unit continuously receives the following data from Patient 1 (a patient with type 2 diabetes): Physiological monitoring data: For three consecutive nights, the continuous glucose monitor recorded a cluster of abnormal events with blood glucose levels below 3.5 mmol / L between 2 and 4 a.m. Medication record data: The patient has been taking gliclazide (a sulfonylurea hypoglycemic drug) for a long time. Patient input status data: The patient recorded "loss of appetite" multiple times within a week through a mobile application, and the home smart scale simultaneously showed that their weight had decreased by 1.5 kg within a week; The association analysis submodule of the inference engine simultaneously identified specific patterns in the above three types of data, which met the pre-set trigger conditions of the association rules (i.e., the simultaneous existence of a specific nocturnal hypoglycemia pattern, evidence of sulfonylurea drug use, and persistent loss of appetite and weight loss).
[0043] Step 12: After the rule is triggered, the inference engine calls the knowledge graph.
[0044] The knowledge graph encodes medical logical pathways such as "decreased renal function → decreased drug clearance → accumulation of sulfonylureas → increased risk of hypoglycemia" and "insufficient nutritional intake → reduced liver glycogen reserves → increased sensitivity to hypoglycemic drugs → increased risk of hypoglycemia." The inference engine module matches and correlates the current data with these pathways, generating two competing clinical hypotheses: "Hypothesis 1: Possibly related to age-related underlying renal insufficiency leading to abnormal drug metabolism"; "Hypothesis 2: Possibly related to recent insufficient nutritional intake."
[0045] Step 13: The inference engine module further calculates the overall risk level of the condition as "high" based on the frequency (three consecutive nights), severity (<3.5 mmol / L), and objectivity and persistence of decreased appetite and weight loss. Subsequently, it outputs the first structured inference sub-result, which includes: a list of identified abnormal patterns, generated clinical hypotheses one and two, and the high-risk level.
[0046] Step 14: Clinical Interaction and Feedback Collection The record summary generation unit automatically fills the above reasoning sub-results into the "Assessment (A)" section of the SOAP medical record format, forming a clinical summary, which is then presented to the physician. After reviewing it, the physician, based on their clinical experience, confirms the possibility of Hypothesis 1 and adds "the possibility of dehydration needs to be investigated," and then confirms the summary.
[0047] Step 15: The clinical interaction optimization unit sends the doctor-confirmed and modified conclusion as clinical validation feedback to the rule optimization module. The rule optimization module performs the following operations: Compare the original output of hypothesis 1 and hypothesis 2 from the system with the doctor's final conclusion (confirm hypothesis 1, supplement with dehydration investigation).
[0048] Locate the association rule that triggered this inference and the data threshold used at that time (such as "weight loss rate threshold").
[0049] Because doctors highly value the "kidney function" association pathway, the rule optimization module can slightly increase the association weight between "weight loss" and "kidney function pathway" in the knowledge graph. Simultaneously, "dehydration" is introduced as a new identifying factor, linked to the logical chain between "kidney function" and "drug accumulation," enhancing the system's future sensitivity to similar cues.
[0050] Example 2: Drug Abnormal Event Discovery and Closed-Loop Optimization Based on Temporal Causality Rules. This example demonstrates how the system can proactively discover and attribute drug side effects using rigorous temporal and pharmacological models.
[0051] Step 21: The data integration and processing unit continuously receives medication records from Patient Two (osteoarthritis) showing the addition of "ibuprofen" (a nonsteroidal anti-inflammatory drug, NSAID). The system marks this new drug event time point T0. Between days 3 and 7 after T0, the system identifies a sustained increase of 12 mmHg in the patient's daily blood pressure monitor readings compared to baseline (consistent with an anomalous trend in the known pressor effect model of NSAIDs). Between days 8 and 10 after T0, the patient reports "ankle swelling" (an associated symptom report). The causal inference submodule identifies this complete time-series event chain, triggering the causal rule.
[0052] Step 22: The inference engine calls the known abnormal causal chain model in the knowledge graph, "NSAID → inhibition of prostaglandin synthesis → renal vasoconstriction → sodium and water retention → increased blood pressure / peripheral edema," for verification: Timing verification: Blood pressure rises first (days 3-7), followed by edema (days 8-10), which is consistent with the physiological evolution of water and sodium retention in the model.
[0053] Clinical manifestations verified that the symptoms (elevated blood pressure, lower extremity edema) were completely consistent with the typical characteristics of the model.
[0054] Meanwhile, the engine actively queried data from the same period and found no new records of other antihypertensive drugs, heart failure, or acute deterioration of kidney function (i.e., no obvious competing events).
[0055] Step 23: Based on the high consistency between the time sequence and clinical manifestations, and the preliminary exclusion of competing events, the system calculates the confidence level of the causal association between this drug and the abnormal event as "high" (e.g., 85%). The second structured inference sub-result is output, including: "Ibuprofen use may have a causal relationship with subsequent new-onset hypertension and lower extremity edema," with a high confidence level.
[0056] Step 24: After receiving this summary, the doctor acknowledges the inference and takes measures to change the medication. The system records that the doctor has confirmed this causal relationship. Upon receiving this confirmation feedback, the rule optimization module can execute: Strengthen the success path: For the causal rule of "ibuprofen" to "elevated blood pressure", appropriately lower the blood pressure elevation threshold required to trigger it or shorten the observation time window to make it more sensitive in future monitoring.
[0057] Knowledge Graph Enhancement: Using the confirmed cases as empirical evidence, the weight of the causal relationship nodes between "ibuprofen" and "water and sodium retention" in the knowledge graph is strengthened.
[0058] In summary, through the workflows demonstrated in Embodiments 1 and 2, this invention embodies its two core innovative mechanisms: proactive reasoning and feedback optimization. The system can not only perform multi-dimensional associative thinking and causal inference based on pharmacological models, like a seasoned physician, but also transform each clinical interaction into nourishment for its own evolution, thereby achieving a transformation from a generalized tool to personalized intelligent assistance.
[0059] 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 preferred examples and are not intended to limit 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 the claimed invention.
Claims
1. A chronic disease management and clinical interaction recording platform for the elderly population, characterized in that, include: A data integration and processing unit is used to access and process raw health data from multiple heterogeneous data sources, and to preprocess the raw health data to generate a fused data sequence. The knowledge base and reasoning unit are used to analyze and reason about the input fused data sequence and output a structured reasoning result containing abnormal patterns and inferred medical correlations. A record summary generation unit is used to receive the structured reasoning results and convert the structured reasoning results into natural language clinical summary text based on a predefined clinical summary template. The clinical interaction optimization unit provides a doctor-side review interface for displaying the clinical summary text and receiving doctor review interaction operations, generating clinical verification feedback and transmitting it to the knowledge base and reasoning unit.
2. The chronic disease management and clinical interaction recording platform for the elderly population according to claim 1, characterized in that: The raw health data includes physiological monitoring data, medication record data, and status data actively entered by elderly patients.
3. The chronic disease management and clinical interaction recording platform for the elderly population according to claim 2, characterized in that: The knowledge base and reasoning unit include a knowledge base module, a reasoning engine module, and a rule optimization module; The knowledge base module is used to store a knowledge graph constructed based on medical knowledge and inference rules defined based on the knowledge graph. The inference rules include at least association rules and causal rules. The inference engine module is used to call the inference rules to analyze the fused data sequence, perform correlation analysis and causal inference, and output structured inference results; The structured reasoning result includes the first structured reasoning sub-result and the second structured reasoning sub-result; The rule optimization module is used to fine-tune the parameters of the historical inference rules that generate the feedback based on the clinical validation feedback from the clinical interaction optimization unit. The parameters include trigger thresholds, association weights, or confidence calculation coefficients.
4. The chronic disease management and clinical interaction recording platform for the elderly population according to claim 3, characterized in that: The knowledge graph encodes medical entities, relationships, and pathophysiological logic related to chronic diseases in the elderly using a graph structure. The association rules and causal rules are defined and invoked based on the graph structure logic of the knowledge graph.
5. The chronic disease management and clinical interaction recording platform for the elderly population according to claim 4, characterized in that: The inference engine module includes an association analysis submodule, which is configured to perform the following steps to implement association rules: S11. When a specific abnormal pattern from physiological monitoring data, a key drug change event from medication record data, and a relevant status indication from patient input status data are identified simultaneously, correlation analysis is triggered. S12. Based on the predefined medical logic paths in the knowledge graph, the specific abnormal patterns, key drug change events and related status indicators are associated to generate clinical hypotheses describing their potential common pathophysiological basis. S13. Assess the overall risk level of the clinical hypothesis based on the severity and frequency of the specific abnormal pattern and the persistence of related state indicators; S14. Output the first structured reasoning sub-result containing the clinical hypothesis and the comprehensive risk level.
6. The chronic disease management and clinical interaction recording platform for the elderly population according to claim 5, characterized in that: The inference engine module includes a causal inference submodule, which is configured to perform the following steps to implement causal rules: S21. When a new drug event is identified in the medication record data, the time of occurrence of the event shall be used as the starting time point. S22. Starting from the aforementioned starting time point, within a preset observation time window, if physiological monitoring data that conforms to the abnormal trend expected by the known abnormal effect causal chain model of the drug in the knowledge graph and symptom reports associated with the patient's input status data are sequentially identified, causal inference is triggered. S23. Call the known abnormal action causal chain model corresponding to the newly added drug in the knowledge graph, verify it and generate a matching score; S24. Proactively query patient data from the same period to identify whether there are competing events that could replace the explanation of the abnormal trends or symptom reports; S25. Based on the matching degree and the exclusion of the competing events, calculate the confidence level of the causal relationship between the new drug event and the subsequent anomaly; S26. Output a second structured inference sub-result containing a description of the causal relationship and the confidence level.
7. The chronic disease management and clinical interaction recording platform for the elderly population according to claim 6, characterized in that: In step S23, the matching score is generated as follows: Determine whether the timing of the abnormal trend and the occurrence of symptom reports matches the expectations of the causal chain model; Determine whether the specific clinical manifestations conform to the typical characteristics described by the causal chain model; Based on the agreement between the two judgments, a time series and clinical feature matching score is generated.
8. The chronic disease management and clinical interaction recording platform for the elderly population according to claim 1, characterized in that: The clinical summary text shall include at least a chronological list of abnormal events, a description of clinical circumstances based on inferred correlations, and itemized questions to be verified.
9. The chronic disease management and clinical interaction recording platform for the elderly population according to claim 1, characterized in that: The doctor's review interface includes a differentiated information display area and an interactive processing area; The differentiated information display area is used to visualize and associate the inferred medical associations with the original health data; the inferred medical associations include clinical hypotheses generated by the association rules or causal relationships inferred by the causal rules; The interactive processing area provides interactive components for checking and confirming or modifying issues to be verified, as well as functions for confirming or editing clinical summary text.