A risk assessment-based inpatient fall prevention and management system

Abstract

This invention provides a risk assessment-based fall prevention and management system for hospitalized patients, relating to the fields of smart healthcare and patient safety technology. It includes a multi-source data fusion module, which collects and processes diverse patient data to form a comprehensive dataset; the multi-source data fusion module includes an IoT sensor array; and a behavioral intent recognition module, which performs collaborative fusion processing on the IoT sensor array to form behavioral intent tags. This risk assessment-based fall prevention and management system for hospitalized patients, through multi-source data fusion and behavioral intent recognition modules, captures real-time changes in the patient's physiological and behavioral states, dynamically assesses fall risk, and forms an accurate dynamic risk profile of the patient. The fall prevention and management system for hospitalized patients overcomes the limitations of traditional static assessment methods, implements precise intervention for high-risk patients, and significantly reduces the probability of falls.

Description

Technical Field

[0001] This invention relates to the field of smart healthcare and patient safety technology, specifically to a fall prevention and management system for hospitalized patients based on risk assessment. Background Technology

[0002] Patient falls are a major challenge in hospital safety management. Falls can lead to secondary injuries such as fractures and traumatic brain injuries, prolonging hospital stays and increasing medical costs. Currently, the most common fall prevention measures in clinical practice rely primarily on manual assessment and routine monitoring. Nursing staff use the Morse Fall Assessment Scale to assess the static risk of patients upon admission. Based on the assessment results, a series of basic preventative measures are implemented for high-risk patients, including displaying warning signs at the bedside, requiring family members to accompany the patient, providing verbal health education to patients and their families, and arranging for patients to be placed in wards near the nurses' station. Nursing staff conduct regular ward rounds to observe patient conditions. Some modern wards are beginning to introduce basic monitoring equipment, such as bed-off alarm mats, which trigger simple audible and visual alarms when a patient's body leaves the mattress to alert medical staff.

[0003] The existing technological system suffers from a fundamental flaw: its risk assessment and early warning mechanisms are static and isolated, failing to achieve continuous, dynamic risk perception and precise, personalized intervention. Current assessments heavily rely on nurses' judgments based on scales at fixed time points, judgments that are subjective and fail to reflect the patient's real-time physiological and behavioral changes. A low-risk patient upon admission may instantly become high-risk due to changes in their condition, medication reactions, or nighttime toilet breaks, but current systems cannot capture this dynamic risk. Simple devices such as bed-leaning alarms only provide a single trigger signal, resulting in a high false alarm rate and an inability to distinguish the intent behind the behavior. This leads to a large waste of nursing resources on invalid alarms, while truly high-risk moments may be missed during rounds. The lack of intelligent decision support based on real-time data makes it difficult to provide accurate early warnings and effective interventions for impending falls. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a risk assessment-based fall prevention and management system for hospitalized patients. The technical problem this invention aims to solve is: how to accurately predict and intervene in the fall risk of hospitalized patients and prevent fall events by generating and implementing dynamic risk assessment and personalized intervention strategies.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a fall prevention and management system for hospitalized patients based on risk assessment, comprising: a multi-source data fusion module, wherein the multi-source data fusion module collects and processes multi-source data from patients to form a comprehensive dataset, and the multi-source data fusion module includes an Internet of Things sensor group.

[0006] The behavioral intent recognition module performs collaborative fusion processing on the IoT sensor group to form behavioral intent labels and rehabilitation training quality assessment results.

[0007] The dynamic risk assessment module dynamically integrates and models the comprehensive dataset, the behavioral intention labels, and the rehabilitation training quality assessment results to form a dynamic risk profile of the patient. The dynamic integration and modeling process includes calculating a comprehensive risk score.

[0008] A personalized strategy generation module analyzes and processes the patient's dynamic risk profile to generate personalized intervention strategies.

[0009] The closed-loop intervention execution module generates intervention execution instructions based on the personalized intervention strategy, uses the patient's behavioral response data after executing the intervention execution instructions as intervention execution effect data, and feeds the intervention execution effect data back to the multi-source data fusion module and the personalized strategy generation module to form a closed-loop control.

[0010] Preferably, the IoT sensor group includes a mattress pressure distribution sensor, a millimeter-wave radar, and an inertial measurement unit. The multi-source data includes electronic medical record data, physiological signals, behavioral signals, and psychological state data. The electronic medical record data includes age, disease diagnosis, medication history, history of falls, and anesthesia information. The physiological signals and behavioral signals are acquired by the IoT sensor group. The psychological state data includes anxiety and depression scores obtained from electronic scales and heart rate variability indicators monitored using wearable devices. The anesthesia information includes the type, dosage, and termination time of anesthetic drugs.

[0011] Preferably, the collaborative fusion processing includes performing fusion analysis on the multi-source signals of the IoT sensor group to extract the patient's multimodal features, and matching the multimodal features with a predefined behavioral sequence library to output the behavioral intent label.

[0012] Preferably, the behavioral sequence library includes high-risk intention behavior sequences and standardized rehabilitation training movement sequences. The high-risk intention behavior sequences determine whether the patient intends to get out of bed. The characteristic conditions for the determination include center of gravity displacement conditions, trunk posture conditions, and lower limb movement conditions. The center of gravity displacement condition is that the patient's center of pressure trajectory continuously moves towards the edge of the bed. The trunk posture condition is that the angle between the patient's trunk and the bed plane exceeds a preset angle. The lower limb movement condition is that the angle between one or both of the patient's lower legs and the bed plane is less than a preset angle. The behavioral intention recognition module obtains real-time sensor data from the Internet of Things sensor group, matches the real-time sensor data with the standardized rehabilitation training movement sequences, and outputs the rehabilitation training quality assessment result. The real-time sensor data includes acceleration, angular velocity, and posture angle features. The matching includes stride consistency, posture angle stability, and movement continuity.

[0013] The preferred steps for dynamic fusion and modeling are as follows: S1. Based on the electronic medical record data in the comprehensive dataset, calculate the static baseline risk value using a preset first quantification rule, and calculate the dynamic behavioral risk value using a preset second quantification rule based on the type and intensity of the behavioral intent tag. S2. The static baseline risk value and the dynamic behavioral risk value are weighted and fused to obtain an initial risk value; S3. Map the rehabilitation training quality assessment result to a rehabilitation adjustment coefficient. The mapping adopts a preset third mapping rule. The rehabilitation adjustment coefficient corrects the initial risk value and outputs the comprehensive risk score. The rehabilitation adjustment coefficient is negatively correlated with the movement stability indicated by the rehabilitation training quality assessment result. S4. Determine the risk level based on the comprehensive risk score, obtain the dominant risk factor based on the contribution of the static baseline risk value, dynamic behavioral risk value and rehabilitation adjustment coefficient, and combine the risk level and the dominant risk factor to output the dynamic risk profile of the patient.

[0014] Preferably, the first quantification rule is a scorecard constructed based on clinical guidelines, the second quantification rule is a lookup table that maps the behavioral intention labels to numerical weights, and the third mapping rule is a piecewise function that maps the rehabilitation training quality score results to a numerical range.

[0015] Preferably, the risk level is determined by the magnitude of the comprehensive risk score, and the risk level includes low, medium and high, and the dominant risk factors include static factors, dynamic behavioral factors, dynamic physiological factors, environmental factors and psychological state factors.

[0016] Preferably, the steps of the analysis and decision processing are as follows: S41. Output the intervention urgency level based on the risk level in the patient's dynamic risk profile; S42. Output the intervention type based on the dominant risk factors in the patient's dynamic risk profile; S43. Generate the personalized intervention strategy based on the intervention urgency and the intervention type.

[0017] The preferred intervention urgency levels include high-risk, medium-risk, and low-risk levels. The high-risk level corresponds to immediate intervention with a response level in seconds. The medium-risk level corresponds to recommended intervention with a response level in minutes. The low-risk level corresponds to routine monitoring. The intervention types include behavioral blocking and assistance, physiological state warning, and environmental adaptive adjustment. When the intervention urgency level is high-risk and the intervention type is blocking and assistance, the personalized intervention strategy is output. The execution targets of the personalized intervention strategy include nurse-side devices, patient-side devices, and environmental devices. The contextual information of the personalized intervention strategy includes whether the current time is night, whether the patient is in the restroom area, the ground humidity level, and the distribution of obstacles.

[0018] Preferably, the intervention execution instructions include sending a warning message to the nurse's mobile terminal, automatically controlling the floor lamp to turn on, and issuing a voice reminder to the patient. The intervention execution effect data includes the confirmation time of the nurse receiving the warning message, the changes in the patient's actions after the floor lamp is turned on, and the patient's behavioral response data after the voice reminder.

[0019] This invention provides a fall prevention and management system for hospitalized patients based on risk assessment. It has the following beneficial effects:

[0020] This invention utilizes multi-source data fusion and a behavioral intent recognition module to capture real-time changes in a patient's physiological and behavioral states, dynamically assess fall risk, and generate an accurate dynamic risk profile of the patient. The inpatient fall prevention and management system overcomes the limitations of traditional static assessment methods, enabling precise intervention for high-risk patients and significantly reducing the probability of falls.

[0021] This risk-assessment-based inpatient fall prevention and management system utilizes a personalized strategy generation module to generate tailored intervention strategies based on each patient's risk level and dominant risk factors. A closed-loop intervention execution module provides real-time feedback on intervention effectiveness, ensuring the effectiveness and timeliness of interventions. The system's implementation has improved patient safety, optimized the allocation of nursing resources, reduced false alarms and ineffective interventions, and enhanced the quality and efficiency of medical services. Attached Figure Description

[0022] Figure 1This is a schematic diagram of the fall prevention and management system for hospitalized patients; Figure 2 This is a schematic diagram of the IoT sensor group and data flow structure; Figure 3 This is a schematic diagram of the behavioral intent recognition process; Figure 4 This is a schematic diagram of the dynamic risk assessment module structure; Figure 5 This is a schematic diagram of the personalized strategy process. Detailed Implementation

[0023] 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.

[0024] Example 1 like Figure 1-5 As shown, this embodiment of the invention provides a fall prevention and management system for hospitalized patients based on risk assessment. It includes a multi-source data fusion module, which collects and processes multi-source data from patients to form a comprehensive dataset. The multi-source data fusion module includes an Internet of Things (IoT) sensor group. The IoT sensor group includes a mattress pressure distribution sensor, millimeter-wave radar, and an inertial measurement unit. The multi-source data includes electronic medical record data, physiological signals, behavioral signals, and psychological state data. The electronic medical record data includes age, disease diagnosis, medication history, previous fall history, and anesthesia information. The physiological and behavioral signals are collected by the IoT sensor group. The psychological state data includes anxiety and depression scores obtained from electronic scales and heart rate variability indicators monitored using wearable devices. The anesthesia information includes the type, dosage, and termination time of anesthetic drugs.

[0025] The behavioral intent recognition module performs collaborative fusion processing on the IoT sensor array to generate behavioral intent labels and rehabilitation training quality assessment results. The fusion processing includes analyzing multi-source signals from the IoT sensor array to extract the patient's multimodal features, and matching these features with a predefined behavioral sequence library to output behavioral intent labels. The behavioral sequence library includes high-risk intention behavior sequences and standardized rehabilitation training movement sequences. High-risk intention behavior sequences determine whether the patient intends to get out of bed; the criteria for this determination include center of gravity displacement, trunk posture, and lower limb movement. Center of gravity displacement is defined as the patient's center of pressure continuously moving towards the bed edge; trunk posture is defined as the angle between the patient's trunk and the bed plane exceeding a preset angle; and lower limb movement is defined as the angle between one or both lower legs and the bed plane being less than a preset angle. The behavioral intent recognition module acquires real-time sensor data from the IoT sensor array and matches this data with the standardized rehabilitation training movement sequences to output rehabilitation training quality assessment results. Real-time sensor data includes acceleration, angular velocity, and posture angle features; the matching criteria include stride consistency, posture angle stability, and movement continuity.

[0026] Mattress pressure distribution sensor: Real-time monitoring of the pressure distribution on the patient on the bed and captures the patient's center of gravity displacement.

[0027] Data example: At a certain moment, the sensor measures the position of the pressure center as follows: =0.45 meters, =0.32 meters, and the center of gravity moves towards the edge of the bed. The system will mark the center of gravity moving towards the edge of the bed as a feature if the displacement of the center of gravity exceeds 0.2 meters, based on a preset threshold.

[0028] Millimeter-wave radar: Used to capture micro-motion features of the human body and extract the movement of the patient's upper and lower limbs. After processing, the signals from the millimeter-wave radar are used to extract action type labels.

[0029] Data example: Sensor signals indicate that the patient has made continuous small movements within a specific time period. If the frequency exceeds 1 Hz and the duration exceeds 3 seconds, the system outputs an activity or slight movement label.

[0030] Inertial measurement unit: measures changes in the patient's posture, such as the angle between the trunk and the bed surface and the angle of the lower limbs.

[0031] Data Example: The measured angle between the patient's torso and the bed surface was θ = 35°, and the angle between the lower leg and the bed surface was... =40°, according to the preset threshold, if θ>30° and <45°, the system marks it as preparing to get out of bed.

[0032] Behavioral sequence library and high-risk behavior judgment: High-risk intentional behavior sequence: Used to identify whether a patient intends to get out of bed. This is determined by matching the patient's behavioral characteristics with a template in the high-risk sequence.

[0033] The characteristic conditions include: Center of gravity displacement condition: If the patient's center of pressure continuously moves towards the edge of the bed, and the displacement exceeds a preset threshold of 0.2 meters, it is determined that the center of gravity has shifted towards the edge of the bed, indicating an intention to get out of bed. Trunk posture condition: If the angle between the trunk and the bed surface exceeds a preset angle of 30°, it is considered that the patient is sitting up or preparing to get out of bed. Lower limb movement condition: If the patient's lower limb angle is less than a preset angle of 45°, accompanied by a change in trunk angle, it indicates that the lower limbs are preparing to bear weight, indicating an intention to get out of bed.

[0034] Standardized rehabilitation training movement sequence: used to assess the quality of a patient's rehabilitation training movements and determine whether they conform to the predetermined standardized movement sequence.

[0035] Data Example: Evaluation is performed by matching features such as step size consistency, attitude angle stability, and motion continuity: Stride length consistency: Whether the patient's left and right stride lengths are consistent during walking, and whether the stride length difference is controlled within a certain range. Postural angle stability: Whether the angle between the patient's trunk and the ground remains stable during rehabilitation training, such as whether the angle fluctuation range is within 5° during static squatting movements. Movement continuity: Whether the rehabilitation movements are smooth, and whether there are any pauses or discontinuous movements.

[0036] Behavioral intent tag generation: Behavioral intent tags are generated by matching them with high-risk behavioral templates and rehabilitation training action templates in a behavioral sequence library. The system will comprehensively analyze the patient's behavioral characteristics and determine whether there is an intention to get out of bed or whether the quality of rehabilitation training is up to standard based on the matching results.

[0037] High-risk behavior intention label: If the center of gravity displacement and posture angle match the characteristics of getting out of bed behavior, the system outputs an intention label for getting out of bed.

[0038] Rehabilitation training quality assessment results: The quality of the patient's rehabilitation training is assessed based on factors such as stride consistency and postural stability. If the results meet the standards, the rehabilitation training quality is considered satisfactory.

[0039] The dynamic risk assessment module dynamically fuses and models comprehensive datasets, behavioral intention labels, and rehabilitation training quality assessment results to form a dynamic risk profile of the patient. This dynamic fusion and modeling process includes calculating a comprehensive risk score. The steps of the dynamic fusion and modeling process are as follows:

[0040] S1. Based on the electronic medical record data in the comprehensive dataset, calculate the static baseline risk value using the preset first quantification rule, and calculate the dynamic behavioral risk value using the preset second quantification rule based on the type and intensity of the behavioral intent label.

[0041] S2. The initial risk value is obtained by weighted fusion of the static baseline risk value and the dynamic behavioral risk value.

[0042] S3. Map the rehabilitation training quality assessment results to a rehabilitation adjustment coefficient. The mapping adopts a preset third mapping rule. The rehabilitation adjustment coefficient corrects the initial risk value and outputs a comprehensive risk score. The rehabilitation adjustment coefficient is negatively correlated with the movement stability indicated by the rehabilitation training quality assessment results.

[0043] S4. Risk level is determined based on comprehensive risk score. Dominant risk factors are obtained based on the contributions of static baseline risk value, dynamic behavioral risk value, and rehabilitation adjustment coefficient. The risk level and dominant risk factors are combined to output a dynamic risk profile of the patient. Risk level is determined by the magnitude of the comprehensive risk score, and risk levels include low, medium, and high. Dominant risk factors include static factors, dynamic behavioral factors, dynamic physiological factors, environmental factors, and psychological state factors.

[0044] The first quantification rule is a scorecard built based on clinical guidelines; the second quantification rule is a lookup table that maps behavioral intention labels to numerical weights; and the third mapping rule is a piecewise function that maps rehabilitation training quality scores to numerical intervals.

[0045] The personalized strategy generation module analyzes and processes the patient's dynamic risk profile to generate personalized intervention strategies. The analysis and decision-making steps are as follows:

[0046] S41. Output the urgency of intervention based on the risk level in the patient's dynamic risk profile.

[0047] S42. Output the intervention type based on the dominant risk factors in the patient's dynamic risk profile.

[0048] S43. Generate personalized intervention strategies based on the urgency and type of intervention.

[0049] Intervention urgency levels are categorized into high-risk, medium-risk, and low-risk levels. High-risk levels correspond to immediate intervention, with a response time of seconds. Medium-risk levels correspond to recommended intervention, with a response time of minutes. Low-risk levels correspond to routine monitoring. Intervention types include behavioral intervention and support, physiological state warning, and environmental adaptive adjustment. When the intervention urgency level is high-risk and the intervention type is intervention and support, a personalized intervention strategy is output. The execution targets of the personalized intervention strategy include nurse-side devices, patient-side devices, and environmental devices. The contextual information for the personalized intervention strategy includes whether it is nighttime, whether the patient is in the restroom area, the ground humidity level, and the distribution of obstacles.

[0050] The closed-loop intervention execution module generates intervention execution instructions based on personalized intervention strategies. It uses patient behavioral response data after executing these instructions as intervention effectiveness data, feeding this data back to the multi-source data fusion module and the personalized strategy generation module to form a closed-loop control system. Intervention execution instructions include sending warning messages to nurses' mobile terminals, automatically turning on floor lights, and issuing voice reminders to patients. Intervention effectiveness data includes the time it takes for nurses to confirm receiving warning messages, changes in patient behavior after turning on floor lights, and patient behavioral responses after receiving voice reminders.

[0051] Example 2 This embodiment calculates static baseline risk value and dynamic behavioral risk value, and combines them with rehabilitation adjustment coefficient to generate a comprehensive risk score, outputting a dynamic risk profile of the patient that includes risk level and dominant risk factors.

[0052] 1. Data Input and Preparation 1.1 Electronic Medical Record Data Electronic medical record data includes age, disease diagnosis, medication history, history of falls, and anesthesia information. It is used to calculate static baseline risk values. For example: Age: 72 years, Disease diagnosis: hypertension, diabetes, History of falls: yes, Anesthesia information: Isoflurane, dosage 2.5 mg / kg, Anesthesia ended 30 minutes prior.

[0053] 1.2 Behavioral Intent Labels The behavioral intent label, output by the behavioral intent recognition module, reflects whether the patient exhibits high-risk behaviors such as the intent to get out of bed. For example: Intent type: intent to get out of bed, intensity: 0.85.

[0054] 1.3 Results of Rehabilitation Training Quality Assessment The assessment of rehabilitation training quality is based on the patient's motor abilities to determine whether their rehabilitation training has been stable and up to standard. For example: Training quality score: 80 out of 100, stride consistency 90%, postural stability 85%.

[0055] 2. Dynamic risk assessment 2.1 Calculation of Static Baseline Risk Value Electronic medical record data is used to calculate static baseline risk values ​​based on a scoring card method using the first quantification rule. The scoring card assigns weights to each data point and calculates the static baseline risk value. For example: age 72 years, assigned a value of 15 points; history of falls, assigned a value of 20 points; diabetes, assigned a value of 10 points; static baseline risk value: 45 points.

[0056] 2.2 Calculation of Dynamic Behavioral Risk Value The type and intensity of behavioral intent labels are used to calculate a dynamic behavioral risk value using a second quantification rule. This rule maps behavioral intent labels to numerical weights, combining this weight with the intensity of the intent to derive the risk value. For example:

[0057] Intention to get out of bed, mapping weight 0.85: 100 represents the standardized risk score.

[0058] 2.3 Initial Risk Value Weighted Fusion The initial risk value is obtained by weighting and fusing the static baseline risk value and the dynamic behavioral risk value. The static baseline risk value has a weight of 0.3, and the dynamic behavioral risk value has a weight of 0.7, calculated as follows:

[0059] The initial risk score is 73.

[0060] 3. Correction of rehabilitation adjustment coefficient The rehabilitation training quality assessment results are mapped to a rehabilitation adjustment coefficient, and the initial risk value is adjusted accordingly. The rehabilitation adjustment coefficient is negatively correlated with the stability of the rehabilitation training quality assessment results. For example, a rehabilitation training quality score of 80 corresponds to a rehabilitation adjustment coefficient of 0.9.

[0061] Revised overall risk score : The revised overall risk score is 65.7.

[0062] 4. Risk Level and Dominant Risk Factor Determination Risk levels are determined based on a comprehensive risk score. The risk level rules are as follows:

[0063] Low risk level: Overall risk score ≤ 40; Medium risk level: 40 < Overall risk score ≤ 70; High risk level: Overall risk score > 70. An overall risk score of 65.7 is classified as medium risk.

[0064] The dominant risk factor is determined based on the contributions of the static baseline risk value, the dynamic behavioral risk value, and the rehabilitation adjustment coefficient. For example, if the static factor contributes 20%, the dynamic behavioral factor contributes 50%, and the rehabilitation adjustment coefficient contributes 30%, then the dominant risk factor is the dynamic behavioral factor.

[0065] The dynamic risk profile includes: Risk level: intermediate, dominant risk factor: dynamic behavioral factor.

[0066] Example 3 This embodiment generates personalized intervention strategies based on the patient's risk level, dominant risk factors, and real-time contextual information, and outputs them to the nursing end, the patient end, and the environment end for execution.

[0067] The personalized strategy generation module receives structured data of the patient's dynamic risk profile, including: Overall risk score: e.g. 65.7, risk level: medium, dominant risk factor: dynamic behavioral factor.

[0068] Context information: Current time: 22:35, Patient location: at the edge of the bed, Environmental conditions: Ground humidity 0.15, slightly damp, Obstacles 0.8 meters away from the patient.

[0069] S41. Output the intervention urgency level based on the risk level in the patient's dynamic risk profile. The urgency of intervention corresponds to the risk level: Low level: routine reminder; Medium level: rapid response; High level: immediate intervention.

[0070] The risk level is medium, and the urgency level for intervention is rapid response.

[0071] S42. Analyze intervention types based on the dominant risk factors in the patient's dynamic risk profile. Intervention types are mapped based on dominant risk factors. The system pre-sets intervention priority strategies for different factors:

[0072] Table 1: Intervention types corresponding to different factors.

[0073] The dominant risk factor is a dynamic behavioral factor, and the intervention type is set as an intervention to prevent getting out of bed.

[0074] S43. Generate personalized intervention strategies based on intervention urgency and type. Personalized strategies are generated based on the urgency and type of intervention, combined with contextual information.

[0075] The system's comprehensive judgment logic is as follows: Nighttime hours: Increase the weighting of patient's weakened proprioception and high fall risk. Patient's position on the edge of the bed: Indicates that the attempt to get out of bed has begun, requiring behavioral intervention. High ambient floor humidity: Automatically trigger environmental devices to prevent slipping. Obstacles are close by: It is recommended that nursing staff promptly check environmental safety.

[0076] Based on the above information, the system outputs a personalized intervention strategy: Intervention strategy examples: Nursing intervention strategy: Send a quick response reminder to the nurse's mobile terminal with the message "Patient showed an intention to get out of bed at 22:35, please check within 3 minutes".

[0077] The nurse's interface indicates: Risk level: Medium, main cause: high risk due to dynamic behavior.

[0078] Patient-side intervention strategy: The voice prompt device plays "Please do not get out of bed. Call the nurse if you need help." The prompt lasts for 5 seconds, and the volume is 45dB to avoid startling the patient.

[0079] Environmental intervention strategy: Automatically turn on the under-bed lamps at 30% brightness to illuminate the area around the bed.

[0080] Additional auxiliary strategy: If the obstacle is ≤1 meter away from the patient, the system will prompt the nurse to check the items 0.8 meters to the left of the patient before entering the room.

[0081] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A risk assessment-based fall prevention and management system for hospitalized patients, characterized in that, include: A multi-source data fusion module, which collects and processes multi-source data from patients to form a comprehensive dataset, includes an Internet of Things (IoT) sensor group. A behavioral intent recognition module, which performs collaborative fusion processing on the IoT sensor group to form behavioral intent labels and rehabilitation training quality assessment results; The dynamic risk assessment module dynamically integrates and models the comprehensive dataset, the behavioral intention tags, and the rehabilitation training quality assessment results to form a dynamic risk profile of the patient. The dynamic integration and modeling process includes calculating a comprehensive risk score. A personalized strategy generation module analyzes and processes the dynamic risk profile of the patient to generate a personalized intervention strategy. The closed-loop intervention execution module generates intervention execution instructions based on the personalized intervention strategy, uses the patient's behavioral response data after executing the intervention execution instructions as intervention execution effect data, and feeds the intervention execution effect data back to the multi-source data fusion module and the personalized strategy generation module to form a closed-loop control.

2. The inpatient fall prevention and management system based on risk assessment according to claim 1, characterized in that: The IoT sensor group includes a mattress pressure distribution sensor, millimeter-wave radar, and an inertial measurement unit. The multi-source data includes electronic medical record data, physiological signals, behavioral signals, and psychological state data. The electronic medical record data includes age, disease diagnosis, medication history, history of falls, and anesthesia information. The physiological signals and behavioral signals are acquired by the IoT sensor group. The psychological state data includes anxiety and depression scores obtained from electronic scales and heart rate variability indicators monitored using wearable devices. The anesthesia information includes the type, dosage, and termination time of anesthetic drugs.

3. The inpatient fall prevention and management system based on risk assessment according to claim 1, characterized in that: The collaborative fusion processing includes fusing and analyzing the multi-source signals of the IoT sensor group to extract the patient's multimodal features, and matching the multimodal features with a predefined behavioral sequence library to output the behavioral intent label.

4. A risk assessment-based inpatient fall prevention and management system according to claim 3, characterized in that: The behavioral sequence library includes high-risk intention behavior sequences and standardized rehabilitation training movement sequences. The high-risk intention behavior sequences determine whether a patient intends to get out of bed. The characteristic conditions for this determination include center of gravity displacement, trunk posture, and lower limb movement. The center of gravity displacement condition is that the patient's center of pressure trajectory continuously moves towards the edge of the bed. The trunk posture condition is that the angle between the patient's trunk and the bed plane exceeds a preset angle. The lower limb movement condition is that the angle between one or both of the patient's lower legs and the bed plane is less than a preset angle. The behavioral intention recognition module acquires real-time sensor data from the IoT sensor group, matches the real-time sensor data with the standardized rehabilitation training movement sequences, and outputs the rehabilitation training quality assessment result. The real-time sensor data includes acceleration, angular velocity, and posture angle features. The matching includes stride consistency, posture angle stability, and movement continuity.

5. A fall prevention and management system for hospitalized patients based on risk assessment according to claim 1, characterized in that: The steps of the dynamic fusion and modeling process are as follows: S1. Based on the electronic medical record data in the comprehensive dataset, calculate the static baseline risk value using a preset first quantification rule, and calculate the dynamic behavioral risk value using a preset second quantification rule based on the type and intensity of the behavioral intent tag. S2. The static baseline risk value and the dynamic behavioral risk value are weighted and fused to obtain an initial risk value; S3. Map the rehabilitation training quality assessment result to a rehabilitation adjustment coefficient. The mapping adopts a preset third mapping rule. The rehabilitation adjustment coefficient corrects the initial risk value and outputs the comprehensive risk score. The rehabilitation adjustment coefficient is negatively correlated with the movement stability indicated by the rehabilitation training quality assessment result. S4. Determine the risk level based on the comprehensive risk score, obtain the dominant risk factor based on the contribution of the static baseline risk value, dynamic behavioral risk value and rehabilitation adjustment coefficient, and combine the risk level and the dominant risk factor to output the dynamic risk profile of the patient.

6. A risk assessment-based inpatient fall prevention and management system according to claim 5, characterized in that: The first quantification rule is a scorecard constructed based on clinical guidelines, the second quantification rule is a lookup table that maps the behavioral intention labels to numerical weights, and the third mapping rule is a piecewise function that maps the rehabilitation training quality score results to a numerical range.

7. A risk assessment-based inpatient fall prevention and management system according to claim 5, characterized in that: The risk level is determined by the magnitude of the comprehensive risk score. The risk level includes low, medium and high. The dominant risk factors include static factors, dynamic behavioral factors, dynamic physiological factors, environmental factors and psychological state factors.

8. A risk assessment-based inpatient fall prevention and management system according to claim 1, characterized in that: The steps of the analysis and decision processing are as follows: S41. Output the intervention urgency level based on the risk level in the patient's dynamic risk profile; S42. Output the intervention type based on the dominant risk factors in the patient's dynamic risk profile; S43. Generate the personalized intervention strategy based on the intervention urgency and the intervention type.

9. A risk assessment-based inpatient fall prevention and management system according to claim 8, characterized in that: The intervention urgency levels include high-risk, medium-risk, and low-risk levels. The high-risk level corresponds to immediate intervention, with a response level in seconds. The medium-risk level corresponds to recommended intervention, with a response level in minutes. The low-risk level corresponds to routine monitoring. The intervention types include behavioral blocking and assistance, physiological state warning, and environmental adaptive adjustment. When the intervention urgency level is high-risk and the intervention type is blocking and assistance, the personalized intervention strategy is output. The execution targets of the personalized intervention strategy include nurse-side devices, patient-side devices, and environmental devices. The contextual information of the personalized intervention strategy includes whether the current time is night, whether the patient is in the restroom area, the ground humidity level, and the distribution of obstacles.

10. A risk assessment-based inpatient fall prevention and management system according to claim 1, characterized in that: The intervention execution instructions include sending a warning message to the nurse's mobile terminal, automatically controlling the floor light to turn on, and issuing a voice reminder to the patient. The intervention execution effect data includes the confirmation time of the nurse receiving the warning message, the changes in the patient's actions after the floor light is turned on, and the patient's behavioral response data after the voice reminder.

Images