Nursing home admission assessment and dynamic resource scheduling system based on health deterioration model

Abstract

The application discloses a nursing home admission evaluation and dynamic resource scheduling system based on a health attenuation model, relates to the technical field of medical resource management, and has the advantages that a multi-source data acquisition and preprocessing module outputs structured time sequence data to a dynamic health portrait generation module for portrait generation, a calculation module calculates a real-time urgency index, a health attenuation analysis and survival risk prediction module adopts an STL algorithm to perform trend decomposition on long-period data, combines a Cox survival model to output a prediction result of a future survival risk probability of the elderly, an admission evaluation and decision module evaluates a residence risk and decides whether to reside, and a task arrangement and scheduling planning module outputs a task queue to be executed through multi-objective optimization. The nursing home admission evaluation and dynamic resource scheduling system based on the health attenuation model improves the work efficiency of the nursing home and reduces the survival risk of the elderly.

Description

Technical Field

[0001] This invention relates to the field of medical resource management technology, and in particular to a nursing home admission assessment and dynamic resource scheduling system based on a health decline model. Background Technology

[0002] Nursing homes in my country generally face the situation of "few medical staff and many disabled elderly people". The existing scheduling model is mostly "fixed room patrol system" (such as patrolling once every 2 hours), which cannot distinguish between key groups and ordinary groups. When an emergency occurs (such as multiple elderly people showing abnormal vital signs at the same time), the system lacks an intelligent priority sorting mechanism. Nurses often handle the cases according to "room number order", which may lead to delays in the treatment of high-risk elderly people.

[0003] Existing systems lack the ability to predict long-term health trends. Traditional vital sign monitoring focuses on "current abnormalities" (such as sudden hypertension) but lacks modeling and analysis of the "frailty trajectory" of the elderly. Institutions cannot scientifically assess the potential risk of major disease development or short-term survival risk before admission, resulting in significant medical disputes and operational risks when admitting high-risk elderly. They are unable to achieve scientific pricing and graded care based on risk levels. Furthermore, there is a lack of unified digital collaboration among doctors (issuing medical orders), nurses (implementing medical orders), and caregivers (providing daily care), leading to delayed information transmission. Summary of the Invention

[0004] The purpose of this invention is to provide a nursing home admission assessment and dynamic resource scheduling system based on a health decline model. The life-cycle health decline model transforms long-term trend changes into triggerable risk precursor signals through STL trend decomposition, slope, acceleration, and inflection point identification. The survival risk assessment model (Cox) outputs the probability prediction results of the elderly's future survival risk, quantifying risk into a decision-making probability output for service admission and operational risk control. A unified task pool and high-performance... Prioritizing response and minimizing path length are the two objectives, and real-time scheduling of task queues is achieved by combining location and floor distribution.

[0005] To achieve the above objectives, this invention provides a nursing home admission assessment and dynamic resource scheduling system based on a health decline model, including a multi-source data acquisition and preprocessing module, a dynamic health profile generation module, and... The module includes: calculation module, health decline analysis and survival risk prediction module, admission assessment and decision-making module, task organization and scheduling planning module, collaborative interaction module, and closed-loop feedback and performance analysis module. Multi-source data acquisition and preprocessing module: responsible for collecting multi-source information such as vital signs data detected by IoT devices, nursing records, doctor's orders, assessment scales, and medical records, completing unit unification, outlier handling, missing value imputation and time alignment, and outputting structured time-series data; Dynamic Health Profile Generation Module: Responsible for creating a digital twin profile for each senior citizen based on structured time-series data, integrating static and dynamic factors to generate a dynamic health profile, and providing core basis for real-time urgency index calculation and survival risk model input; Calculation module: Based on dynamic health profiles, it integrates vital sign deviation, disability level, and past risk coefficients to calculate a real-time urgency index. and will They are mapped to task priorities and response time constraints, serving as the core driving variables for scheduling planning; Health decline analysis and survival risk prediction module: Introduces the concept of vitality trend, uses the STL algorithm to perform trend decomposition on long-term data, tracks the slope and acceleration of long-term trend terms, identifies irreversible negative acceleration inflection points, and combines the Cox survival model to output the probability prediction results of the future survival risk of the elderly. Access Assessment and Decision-Making Module: Responsible for integrating trend inflection signals, survival risk probability, and institutional carrying capacity threshold into a pre-occupancy access risk level system and an in-occupancy urgency level system; Task organization and scheduling planning module: Daily tasks and temporary tasks are put into a unified task pool. Doctor's orders are automatically parsed and broken down into executable nursing task packages. Using genetic algorithm as the core, the task pool is optimized in multiple objectives to minimize task response time and optimize task path, and the output is a task queue to be executed. Collaborative Interaction Module: Build a collaborative system of three terminals: smart medical terminal, smart nursing terminal, and management terminal. The smart medical terminal realizes the issuance of medical orders and task package breakdown, the smart nursing terminal pushes the priority task queue to be executed and puts the exception to the top, and the management terminal displays the backlog of tasks through heat map and triggers cross-regional scheduling suggestions. Closed-loop feedback and performance analysis module: Records the entire time from task assignment to response to completion, analyzes the efficiency of medical staff and the time spent caring for different diseases, and inputs the data results back into the task organization and scheduling planning module to achieve continuous optimization of intelligent task queue generation.

[0006] Preferred, Real-time Urgency Index The specific tasks of the calculation module are as follows: Static factors include age and disability level. Chronic disease severity; dynamic factors include real-time vital signs, IoT alarm levels, and recent mental status assessment; The calculation formula is as follows: ; in, This is the weighting coefficient for the deviation of vital signs. The deviation of vital signs is a comprehensive quantification of real-time vital signs, IoT alarm levels, and recent mental state assessments. This is a weighting coefficient for the disability level. Weighting coefficient of prior risk coefficient, The prior risk coefficient is a comprehensive quantification of age and the severity of chronic diseases; Deviation of vital signs Standardized bias was used for each vital sign. The definition is as follows: ; in, For physical signs The baseline mean is the central value of this sign under normal conditions. For physical signs The baseline standard deviation, To ensure safe bandwidth without triggering penalties, multi-signature deviation adopts a weighted or maximum value strategy; Will Normalized to the range of [0,10] or [0,100] and mapped to task priority and response time limit, serving as hard constraint inputs for the scheduling engine.

[0007] Preferably, the health decline analysis process in the health decline analysis and survival risk prediction module is as follows: The Time Series Decomposition (STL) algorithm is used to decompose long-term indicators into seasonal fluctuations, random noise, and long-term trend terms. The slope of the trend term is tracked to identify irreversible negative acceleration inflection points. The STL trend decomposition process is as follows: For each elderly person, at least one set of key indicator sequences reflecting the body's reserves should be selected. This includes data on blood pressure, blood sugar, weight, and exercise capacity, and preprocessing is performed with uniform temporal granularity, missing value imputation, outlier suppression, and individualized baseline calibration. Preprocessed key indicator sequences The STL is executed, and the specific content is as follows: ; in, For seasonal items, For long-term trend items, This is residual noise.

[0008] Preferably, the STL decomposition parameters are set, the seasonal cycle is determined according to the sampling granularity, the trend smoothing window covers 30–180 days, and robust iteration is enabled to reduce the impact of outliers. The long-term slope is obtained by performing linear regression on the long-term trend term in a sliding window, and the short-term slope is obtained in a short window. The trend acceleration is calculated by the slope difference or second difference. We determine whether the body has entered a period of accelerated decline by using slope thresholds and persistence, whether it is a period of rapid decline in bodily function by using consistency of multiple indicators, and whether it is a hidden high-risk condition by using trend amplitude thresholds.

[0009] Preferably, the specific content of the future survival risk probability prediction results output by the Cox model constructed in the health decline analysis and survival risk prediction module, which combines the multi-dimensional characteristics of the elderly, is as follows: Survival risk events are defined as falls, emergency room visits, hospitalization, and organ failure. Events are marked starting from the date of admission or assessment. The occurrence of an event is marked as 0 or 1, with 1 indicating occurrence and 0 indicating non-occurrence. Design feature vectors By integrating static features and dynamic trend features, a feature system with the same origin as the life cycle health decline model is achieved; It provides quantitative predictions of the probability of survival risk events occurring in the next 90 or 180 days for the elderly, offering core data for risk rating.

[0010] Preferably, the definition of the Cox model is as follows: ; in, For risk rate, For the baseline risk function, The regression coefficients to be estimated are: It is a natural exponential function; Estimate baseline cumulative risk after training The survival function is obtained. : ; Then the future The probability of a survival risk event occurring in a day for: ; in, Choose 90 days or 180 days.

[0011] Preferably, the access assessment and decision-making module includes a pre-occupancy access risk level system and an in-occupancy urgency level system; The pre-acceptance risk level system includes a 5-level risk classification, from low to high: low risk, medium risk, high risk, very high risk, and unacceptable. For low-risk levels, the decision is to grant admission and set a base fee rate. For medium-risk levels, the decision is to grant admission but disclose general risks and set a rate of 1.1 times the base fee rate. For high-risk levels, the decision is to grant admission but require enhanced monitoring and set a rate of 1.3 times the base fee rate or increase the risk deposit. For very high-risk levels, the decision is to grant cautious admission; if admitted, resource allocation will be strengthened and a risk deposit of 1.6 times the base fee rate or higher will be set. For unacceptable levels, the decision is to accept high-risk cases with caution or refer to institutions with stronger capacity. The urgency level system for housing is based on The system is divided into five levels, from low to high: routine, attention, important, critical, and high risk. When a high risk occurs, the system must allow a dynamic replanning mechanism that enables interruption of the current process, immediate handling, and replanning of the path after completion.

[0012] The preferred content of the task organization and scheduling planning module is as follows: The tasks are standardized and described. Each task includes task type, associated elderly person ID, location, estimated time, priority, time window, skill requirements, prerequisites and interruptible attributes. The sources of tasks include periodic routine tasks, IoT alarm triggered tasks, and doctor's order decomposition tasks. A multi-objective optimization design is performed using a genetic algorithm as the core. The objective function is shown below: ; in, This is the priority penalty coefficient. For the mission, Priority For the task The earliest allowed start time, For personnel The execution sequence, This is the actual start time of the task. These are the weighting coefficients that minimize the task response time. Weighting coefficients for optimizing the task path. For path-time functions, Indicates task Whether to assign to personnel, Indicates task Unassigned personnel , Indicates task Assigned to personnel ; Preferably, the set of constraints for the objective function includes unique allocation, non-parallelism, time window satisfaction, skill matching, movement constraints, and priority hard constraints.

[0013] Preferably, the genetic algorithm solution process is as follows: Using chromosome coding representation of task sequences and delimiters, all tasks to be assigned are arranged into sequences, and the sequences are divided into different personnel using delimiters. The initial population is generated using priority insertion and nearest neighbor path heuristics, and the fitness is composed of the objective function and the constraint violation penalty function; Evolutionary computation, swapping, or insertion mutation operators are used to locally improve the task on the same floor using sequential crossover (OX) and mapped crossover (PMX) evolutionary computation, and to enable an elite retention strategy, iterating until the generation G is reached or the improvement is less than a threshold. Stop at this time; When a high-risk emergency task occurs, the task that has already started and cannot be interrupted is locked, and the remaining tasks are rolled back to the task pool. The task is then quickly solved and a new queue is issued using the current location of the medical staff and the remaining tasks as new inputs.

[0014] Therefore, the nursing home admission assessment and dynamic resource scheduling system based on the health decline model described above has the following advantages compared with the prior art: This application ensures high-risk (high The system prioritizes sending abnormal alarms from elderly individuals to nurses, reducing the risk of accidents; it internalizes complex integrated medical and elderly care service processes into standardized tasks automatically distributed by the system, reducing reliance on staff experience; through quantitative prediction of the decline trend of the elderly's life cycle, the system helps institutions identify hidden high-risk groups at the "entry point," avoiding medical disputes caused by blind admission; and it also achieves "risk-based dynamic pricing," ensuring the sound operation of institutions.

[0015] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0016] Figure 1 This is an overall structural diagram of the nursing home admission assessment and dynamic resource scheduling system based on the health decline model of this invention; Figure 2 This is an application interface diagram of the nursing home admission assessment and dynamic resource scheduling system based on the health decline model of this invention. Detailed Implementation

[0017] In the description of this invention, it should be noted that the terms "upper," "lower," "inner," "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of this invention is usually placed when in use. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0018] Example like Figures 1-2 As shown, the nursing home admission assessment and dynamic resource scheduling system based on a health decline model of the present invention includes a multi-source data acquisition and preprocessing module, a dynamic health profile generation module, and The module includes: calculation module, health decline analysis and survival risk prediction module, admission assessment and decision-making module, task organization and scheduling planning module, collaborative interaction module, and closed-loop feedback and performance analysis module. Multi-source data acquisition and preprocessing module: responsible for collecting multi-source information such as vital signs data detected by IoT devices, nursing records, doctor's orders, assessment scales, and medical records, completing unit unification, outlier handling, missing value imputation and time alignment, and outputting structured time-series data; Dynamic Health Profile Generation Module: Responsible for creating a digital twin profile for each senior citizen based on structured time-series data, integrating static and dynamic factors to generate a dynamic health profile, and providing core basis for real-time urgency index calculation and survival risk model input; Calculation module: Based on dynamic health profiles, it integrates vital sign deviation, disability level, and past risk coefficients to calculate a real-time urgency index. and will They are mapped to task priorities and response time constraints, serving as the core driving variables for scheduling planning; The specific job duties are as follows: Static factors include age and disability level. Chronic disease severity; dynamic factors include real-time vital signs, IoT alarm levels, and recent mental status assessment; The calculation formula is as follows: ; in, This is the weighting coefficient for the deviation of vital signs. The deviation of vital signs is a comprehensive quantification of real-time vital signs, IoT alarm levels, and recent mental state assessments. This is a weighting coefficient for the disability level. Weighting coefficient of prior risk coefficient, The Past Risk Coefficient is a comprehensive quantification of age and the severity of chronic diseases (it refers to the cumulative quantification of an elderly person's historical risk, which usually integrates the severity of past medical history, the frequency and severity of risk events that have occurred in the past period, such as falls, emergency room visits, hospitalizations, the types and severity of chronic diseases, etc. It is a static or semi-static factor that quantifies an elderly person's historical health risk experience into a numerical value. The survival risk probability and the severity of chronic diseases output by the Cox model can all be used as sources for this coefficient). Deviation of vital signs Standardized bias was used for each vital sign. The definition is as follows: ; in, For physical signs The baseline mean (or the population normal reference mean) is the central value of this sign under normal conditions. For physical signs The baseline standard deviation (reflecting the dispersion of the normal fluctuations of this sign). To ensure safe bandwidth without triggering penalties, multi-signature deviation adopts a weighted or maximum value strategy; Will Normalized to the range of [0,10] or [0,100] and mapped to task priority and response time limit, serving as hard constraint input for the scheduling engine; Health decline analysis and survival risk prediction module: Introduces the concept of vitality trend, uses the STL algorithm to perform trend decomposition on long-term data, tracks the slope and acceleration of long-term trend terms, identifies irreversible negative acceleration inflection points, and combines the Cox survival model to output the probability prediction results of the future survival risk of the elderly. The health decline analysis process is as follows: The Time Series Decomposition (STL) algorithm is used to decompose long-term indicators into seasonal fluctuations, random noise, and long-term trend terms. The slope of the trend term is tracked to identify irreversible negative acceleration inflection points. The STL trend decomposition process is as follows: For each elderly person, at least one set of key indicator sequences reflecting the body's reserves should be selected. This includes blood pressure, blood sugar, weight, and exercise capacity, and preprocessing is performed with uniform time granularity, missing value imputation, outlier suppression, and individualized baseline calibration. Preprocessed key indicator sequences The STL is executed, and the specific content is as follows: ; in, For seasonal items, For long-term trend items, This is residual noise; Set STL decomposition parameters, with the seasonal cycle determined by the sampling granularity (e.g., daily granularity m=7 / 30), and the trend smoothing window covering 30–180 days. Enable robust iteration to reduce the impact of outliers. Seasonal cycle (m): Determined by the sampling particle size. For example, the daily particle size can be (m=7) (weekly cycle) or (m=30) (monthly cycle), and the hourly particle size can be (m=24). Trend smoothing window: preferably covering 30–180 days (or equivalent number of sampling points) to filter out short-term disturbances; Robust Iteration: Select to enable robust STL and reduce the impact of outliers through residual weight iteration; The long-term slope is obtained by performing linear regression on the long-term trend term in a sliding window, and the short-term slope is obtained in a short window. The trend acceleration is calculated by the slope difference or second difference. On a sliding window (e.g., the most recent 28 / 56 / 90 days) Obtaining the long-term slope by performing linear regression ; Short-term slopes are obtained within a shorter window (e.g., the most recent 7 / 14 days). ; Define acceleration The calculation formula is: Alternatively, the trend acceleration can be obtained using second-order difference. Whether the decline has accelerated is determined by the slope threshold and persistence; whether the body function has declined sharply is determined by the consistency of multiple indicators; and whether the risk is hidden is determined by the trend amplitude threshold. Slope threshold and persistence criterion: If And continue Each sampling point, simultaneously continued If there are 100 sampling points, it is determined that the decay acceleration period has begun; Multi-indicator consistency criteria: If at least two key indicators simultaneously meet the slope threshold and persistence criteria (such as albumin decrease and walking speed decrease) within a certain window, it is determined to be "Frailty Cascade". Trend magnitude threshold criterion: Even if the absolute value does not cross the clinical red line, as long as the trend item decreases relatively by more than the threshold (e.g., ≥20%) within 4–12 weeks, it can be judged as a hidden high risk. The specific details of the Cox model constructed by combining the multi-dimensional characteristics of the elderly and the output of the prediction results of the probability of future survival risk are as follows: Survival risk events are defined as falls, emergency room visits, hospitalization, and organ failure. Events are marked starting from the date of admission or assessment. The occurrence of an event is marked as 0 or 1, with 1 indicating occurrence and 0 indicating non-occurrence. Design feature vectors It integrates static features (age, comorbidity index, historical hospitalization frequency) with dynamic trend features (slope, acceleration, inflection point markers, recent alarm count, etc. of STL decomposition) to achieve a feature system with the same origin as the life cycle health decline model. It provides quantitative output of the predicted probability of survival risk events for the elderly in the next 90 or 180 days, providing core data for risk rating; The Cox model is defined as follows: ; in, For risk rate, For the baseline risk function, The regression coefficients to be estimated are: For the natural exponential function (in terms of the natural constant) Exponential operations with base 0); Estimate baseline cumulative risk after training The survival function is obtained. : ; Then the future The probability of a survival risk event occurring in a day for: ; in, Choose 90 days or 180 days; Access Assessment and Decision-Making Module: Responsible for integrating trend inflection signals, survival risk probability, and institutional carrying capacity threshold into a pre-occupancy access risk level system and an in-occupancy urgency level system; The pre-acceptance risk level system includes a 5-level risk classification, from low to high: low risk, medium risk, high risk, very high risk, and unacceptable. For low-risk levels, the decision is to grant admission and set a base fee rate. For medium-risk levels, the decision is to grant admission but disclose general risks and set a rate of 1.1 times the base fee rate. For high-risk levels, the decision is to grant admission but require enhanced monitoring and set a rate of 1.3 times the base fee rate or increase the risk deposit. For very high-risk levels, the decision is to grant cautious admission; if admitted, resource allocation will be strengthened and a risk deposit of 1.6 times the base fee rate or higher will be set. For unacceptable levels, the decision is to accept high-risk cases with caution or refer to institutions with stronger capacity. For example, the specific details of the classification are shown in Table 1; Table 1. Pre-residence access risk level system and decision-making;

[0019] The urgency level system for housing is based on The system is divided into five levels, from lowest to highest: Routine, Attention, Important, Critical, and High Risk. When a High Risk scenario occurs, the system must allow a dynamic replanning mechanism that enables interruption of the current process, immediate action, and rerouting after completion. For example, [0,10] is shown in Table 2; Table 2. Urgency Level System and Task Response Time in China;

[0020] The task organization and scheduling module: Routine tasks (temperature measurement, medication dispensing, etc.) and ad-hoc tasks (alarm handling, temporary medical orders, etc.) are placed into a unified task pool. Doctors' orders are automatically parsed and broken down into executable nursing task packages. Using a genetic algorithm as the core, the task pool is optimized across multiple objectives to minimize task response time (high efficiency). Prioritize tasks, optimize task paths (combining nurse location and floor distribution to reduce unnecessary running), and output a queue of tasks to be executed; The specific content of the task organization and scheduling planning module is as follows: The tasks are standardized and described, with each task including task type, associated elder ID, location, and estimated time. Priority Time window ( The earliest start time for the task. (Latest task completion time) and skill requirements With prerequisite dependencies (such as measurement, recording, and medication dispensing) and interruptible attributes (the handling of P0 priority tasks cannot be delayed), the task sources cover periodic routine tasks, IoT alarm triggering tasks, and doctor's order decomposition tasks. Decision variables include: , For personnel The execution sequence, This refers to the actual start time of the task. A multi-objective optimization design is performed using a genetic algorithm as the core. The objective function is shown below: ; in, This is the priority penalty coefficient. For the mission, Priority For personnel The execution sequence, This is the actual start time of the task. These are the weighting coefficients that minimize the task response time. Weighting coefficients for optimizing the task path. For path-time functions, Indicates task Whether to assign to personnel, Indicates task Unassigned personnel , Indicates task Assigned to personnel ; The constraint set includes unique allocation, non-parallelism, time window satisfaction, skill matching, movement constraints, and priority hard constraints. The specific details of each constraint are as follows: Unique allocation: ; Cannot be performed concurrently: The time for tasks performed by the same person does not overlap. ; Satisfying the time window: ; In terms of skill matching ,but ; Movement constraints: ,in, This is the actual start time of the next task. This is the actual start time of the previous task. The execution time of the previous task. To the previous mission position To the next task location The start time of the next task must not be earlier than the completion time of the previous task plus the travel time in between. That is, personnel must complete the previous task and reach the next location before they can start the next task. Priority hard constraint: when The forced response delay should not exceed a threshold (e.g., 3–10 minutes), compared to high Consistency must be prioritized; The genetic algorithm solution process is as follows: Using chromosome coding representation of task sequences and delimiters, all tasks to be assigned are arranged into sequences, and the sequences are divided into different personnel using delimiters. The initial population is generated using priority insertion and nearest neighbor path heuristics, and the fitness is composed of the objective function and the constraint violation penalty function; Evolutionary computation, swapping, or insertion mutation operators are used to locally improve the task on the same floor using sequential crossover (OX) and mapped crossover (PMX) evolutionary computation, and to enable an elite retention strategy, iterating until the generation G is reached or the improvement is less than a threshold. Stop at this time; When a high-risk emergency task occurs, the task that has already started and cannot be interrupted is locked, and the remaining tasks are rolled back to the task pool. The task is quickly solved and a new queue is issued using the current location of the medical staff and the remaining tasks as new inputs. Collaborative Interaction Module: Build a collaborative system of three terminals: smart medical terminal, smart nursing terminal, and management terminal. The smart medical terminal realizes the issuance of medical orders and task package breakdown, the smart nursing terminal pushes the priority task queue to be executed and puts the exception to the top, and the management terminal displays the backlog of tasks through heat map and triggers cross-regional scheduling suggestions. Closed-loop feedback and performance analysis module: Records the entire time from task assignment to response to completion, analyzes the efficiency of medical staff and the time spent caring for different diseases, and inputs the data results back into the task organization and scheduling planning module to achieve continuous optimization of intelligent task queue generation.

[0021] Specific Implementation Example 1: Scenario during peak ward rounds at 9:00 AM; Input: The system detected that Grandpa Zhang (Hypertension III) in Room 301 suddenly had his blood pressure rise to 180 / 100; at the same time, Grandma Li in Room 305 needs to have her blood sugar checked as usual. Calculate: Grandpa Zhang's Urgency Index The score surged to 9.5, while Grandma Li's score was 3.0; Dispatch: The system interrupts the nurse's current routine process, pops up a "priority processing" command on the Pad, and navigates to room 301; Action taken: After the nurse finished treating and recording Mr. Zhang's case, the system replanned the route and instructed him to go to room 305 next door.

[0022] Specific Implementation Example 2: Risk Assessment Scenario for Newly Admitted Seniors; Input: Grandpa Li applied for admission and uploaded his medical examination reports and hospital records from the past year; Analysis: The system revealed that although Grandpa Li's blood pressure is currently normal, his albumin level has shown a continuous and stable downward trend over the past 6 months (slope -0.5% / week), and his walking speed has decreased by 30% year-on-year. Warning: The system has determined that Grandpa Li is in the "Frailty Cascade" stage, and predicts an 85% probability of a fall or organ failure within the next 6 months. Decision: The system pushes suggestions to the hospital director: recommend classifying the patient as a specialist care level and signing a "High-Risk Notification Form", or recommend referral to a branch hospital with stronger integrated medical and elderly care capabilities.

[0023] Therefore, this invention employs the aforementioned health decline model-based nursing home access assessment and dynamic resource scheduling system. The life cycle health decline model transforms long-term trend changes into triggerable risk precursor signals through STL trend decomposition, slope, acceleration, and inflection point identification. The survival risk assessment model, Cox, outputs the probability prediction results of the elderly's future survival risk, quantifying the risk into a decision-making probability output for service access and operational risk control. A unified task pool and high-performance... Prioritizing response and minimizing path length are the two objectives, and real-time scheduling of task queues is achieved by combining location and floor distribution.

[0024] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A nursing home admission assessment and dynamic resource scheduling system based on a health decline model, characterized by: Includes a multi-source data acquisition and preprocessing module, a dynamic health profile generation module, The module includes: calculation module, health decline analysis and survival risk prediction module, admission assessment and decision-making module, task organization and scheduling planning module, collaborative interaction module, and closed-loop feedback and performance analysis module. Multi-source data acquisition and preprocessing module: responsible for collecting multi-source information such as vital signs data detected by IoT devices, nursing records, doctor's orders, assessment scales, and medical records, completing unit unification, outlier handling, missing value imputation and time alignment, and outputting structured time-series data; Dynamic Health Profile Generation Module: Responsible for creating a digital twin profile for each senior citizen based on structured time-series data, integrating static and dynamic factors to generate a dynamic health profile, and providing core basis for real-time urgency index calculation and survival risk model input; Calculation module: Based on dynamic health profiles, it integrates vital sign deviation, disability level, and past risk coefficients to calculate a real-time urgency index. and will They are mapped to task priorities and response time constraints, serving as the core driving variables for scheduling planning; Health decline analysis and survival risk prediction module: Introduces the concept of vitality trend, uses the STL algorithm to perform trend decomposition on long-term data, tracks the slope and acceleration of long-term trend terms, identifies irreversible negative acceleration inflection points, and combines the Cox survival model to output the probability prediction results of the future survival risk of the elderly. Access Assessment and Decision-Making Module: Responsible for integrating trend inflection signals, survival risk probability, and institutional carrying capacity threshold into a pre-occupancy access risk level system and an in-occupancy urgency level system; Task organization and scheduling planning module: Daily tasks and temporary tasks are put into a unified task pool. Doctor's orders are automatically parsed and broken down into executable nursing task packages. Using genetic algorithm as the core, the task pool is optimized in multiple objectives to minimize task response time and optimize task path, and the output is a task queue to be executed. Collaborative Interaction Module: Build a collaborative system of three terminals: smart medical terminal, smart nursing terminal, and management terminal. The smart medical terminal realizes the issuance of medical orders and task package breakdown, the smart nursing terminal pushes the priority task queue to be executed and puts the exception to the top, and the management terminal displays the backlog of tasks through heat map and triggers cross-regional scheduling suggestions. Closed-loop feedback and performance analysis module: Records the entire time from task assignment to response to completion, analyzes the efficiency of medical staff and the time spent caring for different diseases, and inputs the data results back into the task organization and scheduling planning module to achieve continuous optimization of intelligent task queue generation.

2. The nursing home admission assessment and dynamic resource scheduling system based on a health decline model according to claim 1, characterized in that: Real-time urgency index The specific tasks of the calculation module are as follows: Static factors include age and disability level. Chronic disease severity; dynamic factors include real-time vital signs, IoT alarm levels, and recent mental status assessment; The calculation formula is as follows: ； in, This is the weighting coefficient for the deviation of vital signs. The deviation of vital signs is a comprehensive quantification of real-time vital signs, IoT alarm levels, and recent mental state assessments. This is a weighting coefficient for the disability level. This is the weighting factor for the historical risk coefficient. The prior risk coefficient is a comprehensive quantification of age and the severity of chronic diseases; Deviation of vital signs Standardized bias was used for each vital sign. The definition is as follows: ； in, For physical signs The baseline mean is the central value of this sign under normal conditions. For physical signs The baseline standard deviation, To ensure safe bandwidth without triggering penalties, multi-signature deviation adopts a weighted or maximum value strategy; Will Normalized to the range of [0,10] or [0,100] and mapped to task priority and response time limit, serving as hard constraint inputs for the scheduling engine.

3. The nursing home admission assessment and dynamic resource scheduling system based on a health decline model according to claim 1, characterized in that: The health decline analysis process in the health decline analysis and survival risk prediction module is as follows: The Time Series Decomposition (STL) algorithm is used to decompose long-term indicators into seasonal fluctuations, random noise, and long-term trend terms. The slope of the trend term is tracked to identify irreversible negative acceleration inflection points. The STL trend decomposition process is as follows: For each elderly person, at least one set of key indicator sequences reflecting the body's reserves should be selected. This includes blood pressure, blood sugar, weight, and exercise capacity, and preprocessing is performed with uniform time granularity, missing value imputation, outlier suppression, and individualized baseline calibration. Preprocessed key indicator sequences The STL is executed, and the specific content is as follows: ； in, For seasonal items, For long-term trend items, This is residual noise.

4. The nursing home admission assessment and dynamic resource scheduling system based on a health decline model according to claim 3, characterized in that: Set STL decomposition parameters, the seasonal cycle is determined by the sampling granularity, the trend smoothing window covers 30–180 days, and enable robust iteration to reduce the impact of outliers. The long-term slope is obtained by performing linear regression on the long-term trend term in a sliding window, and the short-term slope is obtained in a short window. The trend acceleration is calculated by the slope difference or second difference. We determine whether the body has entered a period of accelerated decline by using slope thresholds and persistence, whether it is a period of rapid decline in bodily function by using consistency of multiple indicators, and whether it is a hidden high-risk condition by using trend amplitude thresholds.

5. The nursing home admission assessment and dynamic resource scheduling system based on a health decline model according to claim 1, characterized in that: The specific details of the Cox model constructed by combining multi-dimensional characteristics of the elderly in the health decline analysis and survival risk prediction module are as follows: Survival risk events are defined as falls, emergency room visits, hospitalization, and organ failure. Events are marked starting from the date of admission or assessment. The occurrence of an event is marked as 0 or 1, with 1 indicating occurrence and 0 indicating non-occurrence. Design feature vectors By integrating static features and dynamic trend features, a feature system with the same origin as the life cycle health decline model is achieved; It provides quantitative predictions of the probability of survival risk events occurring in the next 90 or 180 days for the elderly, offering core data for risk rating.

6. The nursing home admission assessment and dynamic resource scheduling system based on a health decline model according to claim 5, characterized in that: The Cox model is defined as follows: ； in, For risk rate, For the baseline risk function, The regression coefficients to be estimated are: It is a natural exponential function; Estimate baseline cumulative risk after training The survival function is obtained. : ； Then the future The probability of a survival risk event occurring in a day for: ； in, Choose 90 days or 180 days.

7. The nursing home admission assessment and dynamic resource scheduling system based on a health decline model according to claim 1, characterized in that: The admission assessment and decision-making module includes a pre-occupancy admission risk level system and an in-occupancy urgency level system; The pre-occupancy access risk level system includes a 5-level risk level classification, from low to high: low risk, medium risk, high risk, ultra-high risk, and uninhabitable. For low-risk level decision-making outputs, the basic fee rate is used for access and pricing. For medium-risk level decision outputs, access is granted but routine risks are disclosed, and the price is 1.1 times the base rate; for high-risk level decision outputs, access is granted but monitoring is strengthened, and the price is 1.3 times the base rate or an additional risk deposit is added; for extremely high-risk level decision outputs, access is granted cautiously, and if access is granted, resource allocation is strengthened and the price is 1.6 times the base rate or a higher risk deposit is added. For institutions with a high risk level that cannot be accommodated, the decision output is "high risk, proceed with caution" or "refer to an institution with a stronger carrying capacity." The urgency level system for housing is based on The system is divided into five levels, from low to high: routine, attention, important, critical, and high risk. When a high risk occurs, the system must allow a dynamic replanning mechanism that interrupts the current process, takes immediate action, and replans the path after completion.

8. The nursing home admission assessment and dynamic resource scheduling system based on a health decline model according to claim 1, characterized in that: The specific content of the task organization and scheduling planning module is as follows: The tasks are standardized and described. Each task includes task type, associated elderly person ID, location, estimated time, priority, time window, skill requirements, prerequisites and interruptible attributes. The sources of tasks include periodic routine tasks, IoT alarm triggered tasks, and doctor's order decomposition tasks. A multi-objective optimization design is performed using a genetic algorithm as the core. The objective function is shown below: ； in, This is the priority penalty coefficient. For the mission, Priority For the task The earliest allowed start time, For personnel The execution sequence, This is the actual start time of the task. These are the weighting coefficients that minimize the task response time. Weighting coefficients for optimizing the task path. For path-time functions, Indicates task Whether to assign to personnel, Indicates task Unassigned personnel , Indicates task Assigned to personnel .

9. The nursing home admission assessment and dynamic resource scheduling system based on a health decline model according to claim 8, characterized in that: The set of constraints for the objective function includes unique allocation, non-parallelism, time window satisfaction, skill matching, movement constraints, and priority hard constraints.

10. The nursing home admission assessment and dynamic resource scheduling system based on a health decline model according to claim 8, characterized in that: The genetic algorithm solution process is as follows: Using chromosome coding representation of task sequences and delimiters, all tasks to be assigned are arranged into sequences, and the sequences are divided into different personnel using delimiters. The initial population is generated using priority insertion and nearest neighbor path heuristics, and the fitness is composed of the objective function and the constraint violation penalty function; Evolutionary computation, swapping, or insertion mutation operators are used to locally improve the task on the same floor using sequential crossover (OX) and mapped crossover (PMX) evolutionary computation, and to enable an elite retention strategy, iterating until the generation G is reached or the improvement is less than a threshold. Stop at this time; When a high-risk emergency task occurs, the task that has already started and cannot be interrupted is locked, and the remaining tasks are rolled back to the task pool. The task is then quickly solved and a new queue is issued using the current location of the medical staff and the remaining tasks as new inputs.

Images