Machine learning-based follow-up timing prediction method, system, medium, and product

Abstract

The machine learning-based follow-up time prediction method, system, medium and product relate to the field of hospital systems. In the method, the application extracts basic information of patients, disease characteristics, treatment response indicators and physiological parameter change trends to form standardized feature vectors. The vectors are input into a preset prediction model to calculate a follow-up risk score and determine the best follow-up time. Then, the follow-up frequency and content are dynamically adjusted to form a personalized follow-up plan. The follow-up task is automatically generated at the predicted time and pushed to relevant medical staff. Finally, the effect data is collected based on the follow-up execution result, and the prediction model parameters are updated. The application changes the follow-up decision from experience judgment to data driving, realizes the precise allocation of follow-up resources, ensures that high-risk patients receive timely intervention, reduces the waste of resources for low-risk patients, and improves the prediction accuracy through continuous optimization of the model, thereby providing scientific and reasonable follow-up decision support for medical institutions.

Description

Technical Field

[0001] This application belongs to the field of hospital systems, and in particular relates to methods, systems, media and products for predicting follow-up timing based on machine learning. Background Technology

[0002] Clinical follow-up is a crucial step in ensuring patients receive continuous treatment and monitoring, playing a key role in chronic disease management and postoperative rehabilitation. However, existing follow-up mechanisms typically employ fixed time intervals, failing to adequately consider individual patient differences and the dynamic nature of disease progression. This leads to irrational allocation of follow-up resources, with high-risk patients not receiving timely follow-up while low-risk patients receive excessive follow-up.

[0003] Several attempts have been made to improve follow-up mechanisms in related technologies, such as standardizing follow-up procedures by developing clinical follow-up guidelines and establishing follow-up reminder systems. Some medical institutions have also begun using electronic health record systems to record patient follow-up information in order to better track patient conditions. Other studies have proposed determining the follow-up frequency for common diseases by analyzing historical data, attempting to establish more scientific follow-up rules.

[0004] However, relying on doctors' experience and judgment to determine the timing of follow-up visits lacks objective and standardized decision-making criteria, resulting in significant discrepancies in follow-up recommendations among different doctors. Furthermore, traditional follow-up models fail to fully utilize patients' electronic health records and real-time monitoring data, making it difficult to achieve precise follow-up management. Summary of the Invention

[0005] This application provides a machine learning-based method, system, medium, and product for predicting follow-up timing, addressing the problems of inaccurate timing determination and unreasonable resource allocation in traditional follow-up mechanisms. This method dynamically predicts follow-up risks and generates personalized follow-up plans by analyzing individual patient characteristics and clinical data, achieving precise allocation of follow-up resources and continuous optimization of follow-up outcomes.

[0006] Firstly, this application provides a machine learning-based method for predicting follow-up timing, including: Obtain patient information, treatment plan data, and historical follow-up records to obtain the raw medical dataset; Based on the original medical dataset, patient basic information, disease characteristics, treatment response indicators, and physiological parameter change trends are extracted to obtain a standardized feature vector. The standardized feature vector is input into a preset prediction model to calculate the patient's follow-up risk score and determine the optimal follow-up time. Based on the follow-up risk score and the optimal follow-up timing, the follow-up frequency and content are dynamically adjusted to obtain a personalized follow-up plan. Based on the personalized follow-up plan, follow-up tasks are automatically generated at the predicted time and pushed to relevant medical staff to obtain follow-up execution results; Based on the follow-up results, follow-up effect data are collected and the parameters of the prediction model are updated to obtain an optimized prediction model.

[0007] In the above embodiments, this application extracts basic patient information, disease characteristics, treatment response indicators, and physiological parameter change trends to form standardized feature vectors. These vectors are then input into a preset prediction model to calculate follow-up risk scores and determine the optimal follow-up timing. The follow-up frequency and content are then dynamically adjusted to form personalized follow-up plans. Follow-up tasks are automatically generated at the predicted timing and pushed to relevant medical staff. Finally, effect data is collected based on the follow-up execution results, and the prediction model parameters are updated. This application transforms follow-up decision-making from experience-based judgment to data-driven approaches, achieving precise allocation of follow-up resources, ensuring timely intervention for high-risk patients, avoiding resource waste for low-risk patients, and continuously improving prediction accuracy through model optimization, providing medical institutions with scientific and reasonable follow-up decision support.

[0008] In one implementation, the step of dynamically adjusting the follow-up frequency and content based on the follow-up risk score and the optimal follow-up timing to obtain a personalized follow-up plan specifically includes: Based on the follow-up risk score, the corresponding follow-up frequency for the patient is determined; Based on the disease characteristics and the treatment response indicators, relevant follow-up items are selected from a preset follow-up content library; Based on the trends in the physiological parameters, key indicators requiring monitoring were identified, and the priority of follow-up content was adjusted accordingly. Based on the prediction results of the optimal follow-up timing, follow-up time points and time windows are generated; The follow-up frequency, follow-up items, key monitoring indicators, and follow-up time are integrated to form a structured personalized follow-up plan.

[0009] In the above embodiments, this application determines personalized follow-up frequency based on follow-up risk scores, realizing intelligent allocation of follow-up resources; selects follow-up items from a preset content library according to disease characteristics and treatment response indicators, improving the effectiveness of clinical monitoring; improves the timeliness of early intervention by dynamically determining key monitoring indicators and adjusting priorities based on physiological parameter change trends; reduces the time cost for both doctors and patients by generating accurate time points and time windows by combining the prediction results of the optimal follow-up timing; and forms a standardized personalized follow-up plan by structurally integrating follow-up frequency, items, indicators, and time, improving the standardization and operability of follow-up implementation.

[0010] In one implementation, the step of automatically generating follow-up tasks and pushing them to relevant medical staff at predicted times according to the personalized follow-up plan, and obtaining follow-up execution results, specifically includes: Based on the time nodes in the personalized follow-up plan, a phased follow-up plan containing multiple follow-ups is generated; Configure the triggering parameters for the follow-up task, including the enrollment time, the most recent discharge time, the last follow-up time, and the time period during which the follow-up can be performed; Acquire real-time clinical data from patients and compare it with the monitoring indicators in the follow-up protocol; Calculate the deviation of follow-up data, and generate a higher priority follow-up task when the monitoring indicator exceeds the preset threshold; The follow-up task will be pushed to the work platform of the medical staff corresponding to the patient; The system receives various indicator data collected during the follow-up process as the follow-up execution results, automatically evaluates the follow-up execution results, and generates a follow-up value trend analysis report.

[0011] In the above embodiments, this application achieves systematic management and long-term continuous monitoring of follow-up tasks by generating phased follow-up plans based on the time nodes of personalized follow-up schemes; it ensures the accuracy of follow-up timing and the continuity of medical processes by configuring multi-dimensional trigger parameters, including enrollment time, discharge time, and last follow-up time; it improves the efficiency of abnormal situation detection by acquiring patients' clinical data in real time and dynamically comparing it with monitoring indicators; it establishes an intelligent risk warning mechanism by calculating the deviation of follow-up data and automatically generating high-priority tasks when the deviation exceeds the threshold; and it improves work efficiency by accurately pushing follow-up tasks to the corresponding medical staff's work platforms.

[0012] In one implementation, the process of generating the personalized follow-up plan further includes: Based on the aforementioned follow-up risk score, assess whether the patient needs to be referred to a higher-level medical institution; When the follow-up risk score exceeds the transfer threshold, the follow-up frequency and follow-up content are automatically adjusted, and preset transfer-related monitoring indicators are added. Calculate the optimal transfer time and target receiving hospital based on the patient's geographical location information and clinical needs; Based on regional medical resource distribution data and real-time traffic information, the optimal transfer route is generated. Based on the optimal transfer timing, the target receiving hospital, the optimal transfer route, the follow-up frequency, and the follow-up content, a personalized follow-up plan including an emergency transfer contingency plan is formed.

[0013] In the above embodiments, this application achieves unified management of routine follow-up and emergency medical care by integrating emergency transfer plans into personalized follow-up plans; establishes an intelligent hierarchical diagnosis and treatment judgment mechanism by automatically assessing referral needs and adjusting monitoring indicators based on risk scores; optimizes medical resource allocation and transfer efficiency by calculating the optimal transfer plan by combining geographical location and clinical needs; and improves the emergency response capability and regional coordination level of medical services by dynamically integrating transfer timing, target hospitals, route planning and follow-up content.

[0014] In one implementation, before pushing the follow-up task to the patient's corresponding medical staff work platform, the method further includes: Obtain resource allocation information and medical staff scheduling information from medical institutions to form a resource availability database; The follow-up tasks in the aforementioned phased follow-up plan are integrated into a preliminary follow-up task list; Based on the resource availability database and the preliminary follow-up task list, the resource load situation in each time period is calculated to obtain a resource load distribution map; Using the resource load distribution map and the time window in the personalized follow-up plan, the specific execution time of the follow-up task is adjusted to generate an optimized follow-up task plan. Based on the optimized follow-up task plan and the professional matching degree of medical staff, follow-up tasks are assigned to the corresponding medical staff to form the final follow-up task.

[0015] In the above embodiments, this application realizes visualized management and load balancing optimization of medical resources by constructing a resource availability database and calculating a resource load distribution map; it reduces resource conflicts and overload and improves resource utilization efficiency by dynamically adjusting task execution time based on resource load and time window; and it improves follow-up quality and work efficiency by combining the professional matching degree of medical staff for task allocation.

[0016] In one implementation, the step of calculating the resource load for each time period based on the resource availability database and the preliminary follow-up task list to obtain a resource load distribution map specifically includes: Extract the number and professional categories of available medical personnel for each time period from the resource availability database; The number of follow-up tasks in each time period of the preliminary follow-up task list is counted, and the professional category required for each task is determined based on the follow-up content. Based on the professional categories required for each follow-up task, calculate the demand for medical staff in different professional categories for each time period; By comparing the demand for medical staff in each time period with the number of available medical staff in the corresponding time period, the resource gap or surplus of each professional category can be obtained. Based on the aforementioned resource gaps or surpluses, calculate the utilization rate of specialized resources and the overall resource balance for each time period; By arranging the utilization rate of professional resources and the overall resource balance of each time period in chronological order, a resource load distribution map in the time dimension is constructed.

[0017] In the above embodiments, this application extracts information on the number of medical staff and their professional categories by time period, establishes an accurate correspondence between tasks and resources by statistically analyzing the number of tasks and matching the needs of professional categories, identifies structural problems in resource allocation by calculating the demand for different professional categories and comparing it with available resources, and improves the scientificity and efficiency of medical resource allocation by quantifying resource utilization and balance and constructing a time-dimensional load distribution map.

[0018] In one implementation, the step of adjusting the specific execution time of the follow-up task using the resource load distribution map and the time window in the personalized follow-up plan to generate an optimized follow-up task plan specifically includes: Identify resource overload periods where resource utilization exceeds a preset threshold and resource idle periods where resource utilization is below a preset threshold from the resource load distribution map; Extract the executable time window and priority score based on clinical needs for each follow-up task from the personalized follow-up plan; Based on the severity of the patient's condition and the degree of abnormality of the monitoring indicators, the urgency score of each follow-up task is calculated, and combined with the priority score, a priority list for task scheduling is determined. According to the priority list, high-priority follow-up tasks will be scheduled during periods with sufficient medical resources within the corresponding executable time window; Identify low-urgency follow-up tasks scheduled during resource overload periods and reschedule them to resource-idle periods, provided that the execution time window is not exceeded. All scheduled follow-up tasks are sorted by execution time to generate an optimized follow-up task plan in a time-series format.

[0019] In the above embodiments, this application achieves dynamic balancing and optimized allocation of medical resource load by identifying resource overload and idle periods; by extracting time windows and priority scores and combining them with the severity of the illness to calculate an urgency score, it prioritizes the allocation of high-priority tasks to periods with sufficient resources, thus ensuring the quality of medical services for critical patients; by rescheduling low-urgency tasks to idle periods, it avoids resource conflicts and overload, thereby improving overall resource utilization efficiency; and by generating time-series optimized task plans, it enhances the intelligence level of follow-up management and the efficiency of medical services.

[0020] In a second aspect, embodiments of this application provide a follow-up timing prediction system based on machine learning, comprising: one or more processors and a memory; the memory is coupled to one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and one or more processors call the computer instructions to cause the system to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a system, cause the system to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, embodiments of this application provide a computer program product that, when run on a system, causes the system to execute the method described in any possible implementation of the first aspect.

[0023] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. This application provides a machine learning-based method for predicting follow-up timing. By extracting basic patient information, disease characteristics, treatment response indicators, and physiological parameter change trends, a standardized feature vector is formed. These vectors are then input into a pre-defined prediction model to calculate a follow-up risk score and determine the optimal follow-up timing. The follow-up frequency and content are then dynamically adjusted to create a personalized follow-up plan. Follow-up tasks are automatically generated at the predicted timing and pushed to relevant medical staff. Finally, effect data is collected based on the follow-up execution results, and the prediction model parameters are updated. This application transforms follow-up decision-making from experience-based judgment to data-driven decision-making, enabling precise allocation of follow-up resources, ensuring timely intervention for high-risk patients, reducing resource waste for low-risk patients, and continuously improving prediction accuracy through model optimization, thus providing medical institutions with scientific and reasonable follow-up decision support.

[0024] 2. This application provides a machine learning-based method for predicting follow-up timing. It determines personalized follow-up frequency based on follow-up risk scores; selects follow-up items from a pre-set content library based on disease characteristics and treatment response indicators; dynamically determines key monitoring indicators and adjusts their priorities based on the changing trends of physiological parameters; generates precise time points and time windows by combining the prediction results of the best follow-up timing; and structurally integrates follow-up frequency, items, indicators, and time to form a standardized personalized follow-up plan, thereby improving the standardization and operability of follow-up execution.

[0025] 3. This application provides a machine learning-based method for predicting follow-up timing. It generates phased follow-up plans based on the time nodes of personalized follow-up protocols, achieving systematic management and long-term continuous monitoring of follow-up tasks. By configuring multi-dimensional trigger parameters, including enrollment time, discharge time, and last follow-up time, it ensures the accuracy of follow-up timing and the continuity of medical processes. By acquiring patient clinical data in real time and dynamically comparing it with monitoring indicators, it improves the efficiency of detecting abnormalities. By calculating the deviation of follow-up data and automatically generating high-priority tasks when it exceeds a threshold, it establishes an intelligent risk warning mechanism. By accurately pushing follow-up tasks to the corresponding medical staff's work platforms, it improves work efficiency. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a follow-up timing prediction method based on machine learning in an embodiment of this application.

[0027] Figure 2 This is another flowchart illustrating a follow-up timing prediction method based on machine learning in an embodiment of this application.

[0028] Figure 3 This is a schematic diagram of the physical device structure of a follow-up timing prediction system based on machine learning provided in an embodiment of this application. Detailed Implementation

[0029] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0030] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0031] In the field of healthcare management, clinical follow-up is a crucial link in ensuring patients receive continuous treatment and monitoring, playing a key role in chronic disease management and postoperative rehabilitation. With the continuous improvement of medical informatization and the ongoing expansion of the chronic disease patient population, how to achieve scientific and rational follow-up management has become a significant challenge for medical institutions.

[0032] In related technologies, existing follow-up mechanisms typically employ fixed time intervals, failing to adequately consider individual patient differences and the dynamic nature of disease progression. This leads to unreasonable allocation of follow-up resources, with high-risk patients not receiving timely follow-up and low-risk patients receiving excessive follow-up. While some attempts have been made to improve follow-up mechanisms, such as standardizing follow-up procedures by developing clinical follow-up guidelines and establishing follow-up reminder systems, and some medical institutions have begun using electronic health record systems to record patient follow-up information, and some studies have proposed determining the follow-up frequency for common diseases by analyzing historical data, these methods still rely on doctors' experience and judgment to determine follow-up timing. They lack objective and standardized decision-making criteria, resulting in significant differences in follow-up recommendations among different doctors. Furthermore, they fail to fully utilize patients' electronic health records and real-time monitoring data, making it difficult to achieve precise follow-up management.

[0033] This application is primarily applied to scenarios such as hospital outpatient follow-up management, long-term monitoring of chronic diseases, postoperative rehabilitation tracking, and post-discharge management of critically ill patients. It is particularly suitable for disease types requiring long-term continuous monitoring, such as cardiovascular diseases, diabetes, and cancer. In these application scenarios, patients' disease status, treatment response, and risk levels exhibit significant individual differences and dynamic changes, making traditional fixed follow-up models insufficient to meet the needs of precision medical services. To address the aforementioned technical problems, this application provides a follow-up timing prediction method, system, medium, and product based on machine learning. An embodiment is described below, combined with… Figure 1 The present application describes a follow-up timing prediction method based on machine learning in its embodiments: Please see Figure 1 This is a flowchart illustrating a follow-up timing prediction method based on machine learning in an embodiment of this application.

[0034] S101. Obtain the patient's condition information, treatment plan data, and historical follow-up records to obtain the original medical dataset.

[0035] Specifically, the system first connects to the hospital's electronic medical record system, laboratory information system, and follow-up management system via API interfaces to automatically retrieve patients' basic information and clinical data. For disease information, the system extracts structured data such as the patient's primary and secondary diagnoses, disease progression description, and symptom scores, while simultaneously using natural language processing technology to parse unstructured text information from doctors' medical records. Acquiring treatment plan data includes extracting medication types, dosages, and frequencies from the medication management system, and obtaining key information such as surgery type, time, and outcome from surgical records.

[0036] S102. Based on the original medical dataset, extract basic patient information, disease characteristics, treatment response indicators, and physiological parameter change trends to obtain standardized feature vectors.

[0037] Standardized feature vectors transform a patient's multidimensional medical data into a numerical vector form that can be processed by machine learning algorithms. Basic patient information includes demographic characteristics such as age, gender, and body mass index; disease characteristics encompass clinical features such as disease type, severity, and duration; treatment response indicators reflect the patient's responsiveness to the current treatment plan; and physiological parameter trends describe the patterns of change in key physiological indicators over time. Standardization ensures that features of different dimensions can be compared and calculated on the same scale.

[0038] Specifically, the system first categorizes and encodes basic patient information, converting categorical variables such as gender into numerical forms and normalizing continuous variables such as age and BMI. Disease features are extracted by converting disease diagnoses into standardized codes and calculating disease severity indicators using a disease severity rating scale. Treatment response indicators are calculated based on the rate of change of key indicators before and after treatment, such as the magnitude of blood pressure reduction and blood glucose control. Physiological parameter trends are analyzed using a sliding window time series analysis method, employing a linear regression algorithm to calculate the slope of changes in various physiological indicators. Statistical methods are used to quantify volatility, and outlier frequencies are identified based on statistical thresholds, forming a trend feature vector reflecting the dynamic changes in the patient's condition.

[0039] In one alternative implementation, principal component analysis (PCA) can be used to reduce the dimensionality of the high-dimensional feature vectors, retaining the most important feature components. Another approach is to use domain-knowledge-based feature selection algorithms to select the most predictive subset of features based on clinical experience and statistical significance. A dynamic feature weight adjustment mechanism can also be implemented to adaptively adjust the importance weights of each feature according to different disease types and patient groups.

[0040] S103. Input the standardized feature vector into the preset prediction model, calculate the patient's follow-up risk score, and determine the optimal follow-up time.

[0041] The pre-defined prediction model employs an ensemble learning architecture, including gradient boosting decision trees, random forests, and logistic regression, fusing prediction results through weighted voting. Model training utilizes historical data from no fewer than 10,000 patient samples, with 70% used for training, 20% for validation, and 10% for testing. The follow-up risk score is calculated as α × disease severity score + β × treatment response index + γ × physiological parameter abnormality + δ × historical follow-up interval deviation, where the weight parameters α, β, γ, and δ are determined through grid search optimization, satisfying α + β + γ + δ = 1. The optimal follow-up timing is determined using a multi-objective optimization algorithm, with the objective function being the minimum of medical costs + risk losses - health benefits, constrained by conditions including clinical safety interval ≤ T ≤ resource availability time window.

[0042] Specifically, the system inputs standardized feature vectors into the ensemble learning model. The model first calculates the patient's baseline risk score using the risk prediction submodule, and then analyzes the optimal follow-up time window using the timing prediction submodule. The risk score is calculated based on the probability of adverse events in similar patients in historical data, combined with a weighted average of the current patient's feature similarity. The optimal follow-up timing is determined using a dynamic programming algorithm to find the most cost-effective time point while meeting clinical safety requirements.

[0043] In one alternative implementation, deep learning models such as LSTM networks can be used to process time-series medical data to better capture the dynamic trends in a patient's condition. Another approach is to use Bayesian network models, which can handle complex dependencies between features and provide uncertainty quantification.

[0044] In some embodiments, the model may make inaccurate predictions for certain rare diseases or specific patient groups. In such cases, when the prediction confidence level falls below a threshold, the system will mark the case as requiring manual review by an expert. To address the lack of historical data for new patients, the system employs a cold-start strategy, providing initial predictions based on statistical patterns of similar patient groups and gradually optimizing as data accumulates. When encountering sudden changes in clinical condition, the system provides an emergency reassessment function, updating risk scores and follow-up timing in real time.

[0045] S104. Based on the follow-up risk score and the optimal follow-up time, dynamically adjust the follow-up frequency and content to obtain a personalized follow-up plan.

[0046] The personalized follow-up plan is a customized follow-up plan based on each patient's specific situation, including elements such as follow-up frequency, follow-up content, and monitoring focus. Follow-up frequency refers to the time interval between two consecutive follow-ups, with higher frequencies for high-risk patients; follow-up content refers to the list of examinations and assessments required for each follow-up, determined based on the patient's disease characteristics and treatment stage; dynamic adjustment means that as the patient's condition changes and treatment progresses, the various parameters of the follow-up plan are adjusted in real time to ensure the timeliness and applicability of the plan.

[0047] Specifically, the system determines the corresponding follow-up frequency for patients based on follow-up risk scores; selects relevant follow-up items from a pre-set follow-up content library based on disease characteristics and treatment response indicators; identifies key monitoring indicators based on physiological parameter change trends and adjusts the priority of follow-up content; generates follow-up time points and time windows by combining the prediction results of the optimal follow-up timing; and integrates follow-up frequency, follow-up items, key monitoring indicators, and follow-up time to form a structured, personalized follow-up plan. Furthermore, the system establishes a segmented mapping relationship between risk scores and follow-up intervals: 30-90 days for low-risk patients, 7-30 days for medium-risk patients, and 1-7 days for high-risk patients, with adjustments made based on factors such as disease acuteness / chronicity and patient age. The follow-up content library is divided into three levels: basic monitoring items, disease-specific items, and risk-oriented items. The system automatically matches corresponding items based on the patient's disease code and individualized risk factors. Key monitoring indicators identify abnormally changing physiological parameters through time series analysis; when the rate of change of an indicator exceeds a preset threshold, it is automatically included in the key monitoring scope. The follow-up time points are established based on a probability distribution model of the disease's periodic characteristics and the treatment's effect cycle, setting key follow-up nodes and allowing for flexible adjustments.

[0048] In one alternative implementation, a tiered follow-up strategy can be established, providing remote patients with a hybrid follow-up model combining home monitoring with regular face-to-face consultations, based on the patient's geographical location and accessibility to medical care. Another approach is to implement a collaborative follow-up mechanism, arranging joint follow-ups by multidisciplinary teams based on the patient's complexity and specialist needs. Adaptive learning algorithms can also be employed to continuously optimize the rules for generating personalized treatment plans based on feedback from follow-up outcomes.

[0049] In some embodiments, patients may be unable to attend follow-up appointments at the scheduled time or experience sudden changes in their condition. In such cases, the system allows adjustments to follow-up times within a certain timeframe and corresponding adjustments to subsequent follow-up plans. For patients with rapidly changing conditions, the system provides an emergency follow-up trigger function, automatically initiating an expedited follow-up process when monitored indicators exceed safe thresholds. When patient compliance is poor, the system adjusts the follow-up methods and content, adding soft interventions such as health education and psychological support.

[0050] S105. Based on the personalized follow-up plan, automatically generate follow-up tasks at the predicted time and push them to relevant medical staff to obtain the follow-up execution results.

[0051] The follow-up task is a specific work instruction generated based on a personalized follow-up plan, containing detailed information such as patient information, follow-up content, execution time, and responsible personnel. Automatic generation means that the system can create corresponding tasks without manual intervention based on preset time trigger conditions and business rules.

[0052] Specifically, the system generates a phased follow-up plan containing multiple follow-ups based on the time nodes in the personalized follow-up plan; configures the trigger parameters for follow-up tasks, including enrollment time, most recent discharge time, last follow-up time, and the executable follow-up time period; acquires real-time clinical data of patients and compares it with the monitoring indicators in the follow-up plan; calculates the deviation of follow-up data, and generates a higher-priority follow-up task when the monitoring indicator exceeds a preset threshold; pushes the follow-up task to the work platform of the corresponding medical staff; receives various indicator data collected during the follow-up process as the follow-up execution result, and automatically evaluates the follow-up execution result to generate a follow-up value trend analysis report. The phased follow-up plan is generated using a sliding window algorithm. The system uses the current time as the starting point and extrapolates all follow-up nodes within the next 3-6 months to form a task sequence on the timeline. The trigger parameter configuration is based on multi-dimensional time calculations. Enrollment time serves as the baseline time point, most recent discharge time is used to calculate the postoperative follow-up interval, the last follow-up time determines the earliest executable time for the next follow-up, and the executable time period considers the working time of medical staff and the patient's cooperation time. Deviation of follow-up data is calculated using a standardized Z-score algorithm. A medium-priority task is triggered when the absolute Z-score is greater than 2, and a high-priority task is triggered when it is greater than 3. Task assignment adopts a role-based distribution mechanism: attending physicians receive comprehensive follow-up tasks, specialist nurses receive specialized monitoring tasks, and rehabilitation therapists receive functional assessment tasks. The automatic evaluation of follow-up results includes data integrity checks, outlier identification, and trend change analysis, generating a visual report containing key indicator change curves and early warning prompts.

[0053] In one alternative implementation, intelligent task prioritization can be implemented, automatically adjusting the task execution order based on patient risk level, task urgency, and healthcare staff expertise. Another approach is to establish a task collaboration mechanism, where complex follow-up tasks can be broken down into multiple sub-tasks and assigned to healthcare professionals from different specialties for collaborative completion.

[0054] In some embodiments, there may be situations where medical staff are temporarily unable to perform tasks or patients are unable to cooperate with follow-up visits. In such cases, when the originally assigned staff cannot respond in a timely manner, the system automatically transfers the task to a backup staff member. For cases where patients are unreachable or refuse follow-up visits, the system will mark them as special circumstances and initiate a multi-channel contact process, including telephone follow-ups, SMS reminders, and contact with family members.

[0055] S106. Based on the follow-up results, collect follow-up effect data and update the parameters of the prediction model to obtain the optimized prediction model.

[0056] In this embodiment, follow-up effectiveness data refers to quantitative indicators of follow-up quality and clinical value obtained by analyzing follow-up execution results, including follow-up timeliness, patient satisfaction, disease control effectiveness, and adverse event incidence. Model parameter updates refer to adjusting the internal parameters of the machine learning model, such as weights and thresholds, based on new data feedback, enabling the model to continuously learn and improve its predictive performance.

[0057] In one embodiment, the process of generating a personalized follow-up plan further includes: assessing whether the patient needs to be transferred to a higher-level medical institution based on the follow-up risk score; automatically adjusting the follow-up frequency and content and adding preset transfer-related monitoring indicators when the follow-up risk score exceeds the transfer threshold; calculating the optimal transfer time and target receiving hospital based on the patient's geographical location information and clinical needs; generating the optimal transfer route based on regional medical resource distribution data and real-time traffic information; and forming a personalized follow-up plan that includes an emergency transfer plan based on the optimal transfer time, target receiving hospital, optimal transfer route, follow-up frequency, and follow-up content.

[0058] The transfer threshold is a risk score cutoff value set based on the patient's disease type and the current medical institution's treatment capacity. When a patient's risk score exceeds this threshold, it indicates that existing medical resources cannot meet the patient's safe treatment needs. Transfer-related monitoring indicators include early warning indicators of disease deterioration and vital sign stability indicators during transfer. The optimal transfer timing is the best transfer time window determined by comprehensively considering factors such as patient disease progression prediction, the availability of resources at the receiving hospital, and transportation conditions.

[0059] Specifically, the system sets tiered transfer thresholds. For example, a patient's risk score exceeding 60 in a primary hospital triggers a transfer assessment at a secondary hospital, and exceeding 80 triggers a transfer assessment at a tertiary hospital; a patient's risk score exceeding 85 in a secondary hospital triggers a transfer assessment at a tertiary hospital. Transfer-related monitoring indicators have been expanded beyond existing follow-up content to include transfer risk assessment items such as heart rate variability, blood oxygen saturation change rate, and level of consciousness score. The monitoring frequency has been adjusted to once every 2-4 hours. The optimal transfer timing is calculated using a multi-objective optimization algorithm, comprehensively evaluating factors such as the urgency of the patient's condition, the availability of beds at the receiving hospital, accessibility, and weather conditions. Target hospital selection is based on an evaluation model established using four dimensions: specialty matching, distance, historical transfer success rate, and current receiving capacity. Transfer route planning integrates real-time traffic data and ambulance GPS positioning, employing a dynamic path optimization algorithm to calculate the route with the shortest estimated arrival time.

[0060] Optionally, the system can establish a regional transfer coordination center to achieve unified scheduling and information sharing of transfer resources among multiple hospitals. Another approach is to integrate remote consultation functionality, using video consultations to confirm the necessity of transfer and the receiving hospital's readiness before the transfer.

[0061] In the above embodiments, by establishing a complete machine learning-driven follow-up prediction system, the entire process of automated management—from data acquisition, feature extraction, model prediction, plan generation, task execution to effect feedback—is achieved. Feature engineering based on multi-dimensional medical data improves prediction accuracy, enabling accurate identification of individualized patient follow-up needs; a dynamic adjustment mechanism ensures that the follow-up plan is synchronized with changes in the patient's condition; automated task distribution and execution result evaluation improve medical work efficiency and reduce human error; and continuous learning and model optimization mechanisms guarantee the long-term effectiveness and adaptability of the system.

[0062] In the above embodiments, the system can automatically generate personalized follow-up plans and push follow-up tasks based on predictive models, effectively improving the intelligence level of follow-up management. However, in actual medical environments, the limited medical resources and the uneven workload of medical staff may lead to follow-up tasks not being executed in a timely and effective manner, affecting the overall follow-up quality. To further optimize resource allocation efficiency and ensure the reasonable allocation and smooth execution of follow-up tasks, this application also provides another follow-up timing prediction method based on machine learning. The following is combined with... Figure 2 Another follow-up timing prediction method based on machine learning is described in the embodiments of this application: Please see Figure 2 This is another flowchart illustrating a follow-up timing prediction method based on machine learning in an embodiment of this application.

[0063] Before pushing the follow-up task to the work platform of the corresponding medical staff for the patient, the method also includes: S201. Obtain resource allocation information and medical staff scheduling information from medical institutions to form a resource availability database.

[0064] Resource allocation information refers to the basic configuration data of various medical resources within a medical institution, including static information such as human resource structure, equipment configuration status, and department setup. Staff scheduling information covers dynamic time-related information such as staff work hours, shift rotations, leave plans, and training arrangements.

[0065] Specifically, the system first automatically obtains the organizational structure information of the medical institution through API interfaces with the hospital's information system and human resource management system, including basic configuration data such as staffing levels, equipment lists, and bed numbers for each department. Subsequently, the system connects to the scheduling management system to obtain real-time dynamic information such as work schedules, on-call arrangements, and leave records for each medical staff member. The system categorizes and organizes the acquired data according to time and professional categories, establishing a multi-dimensional resource index structure. Each medical staff member's information record includes fields such as employee ID, name, professional category, skill level, current work status, and available time slots. The system adopts a multi-level time slot management mechanism, using different time granularities based on task type. For example, routine follow-up tasks use a 30-minute basic time slot, complex assessment tasks use a 60-minute time slot, and urgent follow-up tasks are handled in an immediate response mode. Each time slot has a 20% buffer time to accommodate time deviations during actual execution. The system establishes a dynamic load monitoring mechanism; when the expected workload within a time slot exceeds 80% of the standard workload, a load balancing algorithm is automatically triggered to adjust deferred tasks to adjacent time slots. In the event of an emergency medical situation, the system provides an emergency queue-jumping function, which automatically reschedules tasks for the affected time period.

[0066] S202. Integrate the follow-up tasks in the phased follow-up plan into a preliminary follow-up task list.

[0067] The phased follow-up plan refers to a phased follow-up arrangement based on the patient's disease characteristics and recovery progress, with each phase including different monitoring focuses and intervention measures. The follow-up task is the smallest execution unit in the phased follow-up plan, with clearly defined execution content, time requirements, and resource needs. The preliminary follow-up task list is a set of tasks to be scheduled formed by standardizing and organizing the scattered follow-up tasks of multiple patients according to a unified format.

[0068] Specifically, the system extracts all follow-up tasks to be executed from the personalized follow-up plan database and categorizes them according to patient ID, task type, execution time, and other dimensions. For each follow-up task, the system automatically parses the task description text to identify the required professional skills and the estimated execution time.

[0069] In one alternative implementation, the system can use natural language processing technology to automatically identify key information in the task description, reducing the workload of manual configuration. Another approach is to dynamically adjust the estimated execution time and resource requirements of the task based on its complexity and the severity of the patient's condition. Task clustering analysis can also be implemented to merge similar tasks for similar patients, improving execution efficiency.

[0070] S203. Based on the resource availability database and the preliminary follow-up task list, calculate the resource load for each time period to obtain the resource load distribution map.

[0071] Resource load refers to the comparison between the demand and availability of medical resources within a specific time period, reflecting the degree of resource strain and balance. A resource load distribution map is a chart that visualizes resource load over time.

[0072] Specifically, the system extracts the number of available medical staff and their professional categories for each time period from the resource availability database; it counts the number of follow-up tasks for each time period in the preliminary follow-up task list and determines the required professional category for each task based on the follow-up content; it calculates the demand for medical staff in different professional categories for each time period based on the professional category required for each follow-up task; and it compares the demand for medical staff in each time period with the number of available medical staff in the corresponding time period to obtain the resource gap or surplus for each professional category. For example, the system establishes a task-professional mapping relationship, mapping nursing tasks such as blood pressure monitoring and medication guidance to the nursing profession, and medical tasks such as cardiac function assessment and rehabilitation guidance to the corresponding specialist doctors; it counts the task execution time requirements for each professional category in 15-minute time slots and converts them into the number of personnel required based on standard work efficiency; it calculates the resource load rate, determines it through a weighted calculation method, comprehensively considers the task execution time and complexity coefficient, and compares it with the working capacity of available medical staff. The system establishes a hierarchical load assessment system, using different colors to indicate the load level. When the load exceeds a preset threshold, the resource allocation mechanism is automatically triggered. Finally, a load distribution map in the form of a heat map is generated, which intuitively displays the resource tension of each profession at different times with the light and dark colors.

[0073] In one alternative implementation, the system predicts resource load trends for future periods based on historical load data and external influencing factors.

[0074] In some embodiments, there may be sudden large-scale follow-up needs or large-scale leave of medical staff, which may lead to inaccurate load calculation. In response, the system establishes an abnormal load detection algorithm. When the load exceeds twice the standard deviation of the historical average for a certain period of time, an early warning is triggered and an emergency dispatch plan is automatically activated.

[0075] S204. Using the resource load distribution map and the time window in the personalized follow-up plan, adjust the specific execution time of the follow-up task to generate an optimized follow-up task plan.

[0076] The time window is the acceptable execution time range set for each follow-up task in the personalized follow-up plan, determined comprehensively based on clinical needs, patient convenience, and medical safety. Task adjustment refers to the optimization process of rescheduling follow-up tasks from resource-constrained periods to resource-sufficient periods under the constraint of the time window.

[0077] Specifically, the process involves identifying resource overload periods where resource utilization exceeds a preset threshold and resource idle periods where resource utilization is below a preset threshold from the resource load distribution map; extracting the executable time window and priority score based on clinical needs for each follow-up task from the personalized follow-up plan; calculating the urgency score for each follow-up task based on the severity of the patient's condition and the degree of abnormality in monitoring indicators, and determining the priority order list for task scheduling based on the priority score; prioritizing high-priority follow-up tasks within the corresponding executable time window during periods with sufficient medical resources, for example, using a multi-constraint optimization algorithm to find the globally optimal task allocation scheme under constraints such as time window, resource capacity, and professional skill matching, achieving a balance between maximizing medical resource utilization and minimizing patient waiting time; finally, identifying low-urgency follow-up tasks scheduled during resource overload periods and rescheduling them to resource idle periods while ensuring they do not exceed the executable time window; and sorting all scheduled follow-up tasks by execution time to generate a time-series optimized follow-up task plan.

[0078] S205. Based on the optimized follow-up task plan and the professional matching degree of medical staff, the follow-up tasks are assigned to the corresponding medical staff to form the final follow-up tasks.

[0079] Specifically, the system first establishes a competency evaluation model for medical staff, extracting basic information such as each staff member's educational background, professional certifications, and years of service from their personnel files. This information is then combined with historical task completion records, patient satisfaction ratings, and peer reviews to calculate a comprehensive competency score. For different types of follow-up tasks, the system sets corresponding weights for professional skills requirements. For example, blood glucose monitoring tasks place greater emphasis on nursing skills and communication abilities, while cardiac rehabilitation assessment tasks emphasize specialized medical knowledge and clinical experience.

[0080] In the above embodiments, dynamic monitoring and unified management of medical resources were achieved by establishing a resource availability database, scattered follow-up tasks were standardized and integrated to form a schedulable task set, resource bottlenecks were identified based on the visualization analysis of the load distribution map, intelligent scheduling optimization of tasks was completed by combining time window constraints and priority evaluation, and finally the optimal matching of tasks and personnel was achieved through professional matching degree evaluation.

[0081] The system in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link / reference needed]. Figure 3 This is a schematic diagram of the physical device structure of a follow-up timing prediction system based on machine learning, provided in an embodiment of this application.

[0082] It should be noted that, Figure 3 The structure of the system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0083] like Figure 3 As shown, the system includes a CPU 301, which can perform various appropriate actions and processes according to a program stored in ROM 302 or a program loaded from storage section 308 into RAM 303, such as executing the methods described in the above embodiments. RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via bus 304. I / O interface 305 is also connected to bus 304.

[0084] The following components are connected to I / O interface 305: input section 306 including a camera, infrared sensor, etc.; output section 307 including a liquid crystal display (LCD) and speakers, etc.; storage section 308 including a hard disk, etc.; and communication section 309 including a network interface card such as a LAN (Local Area Network) card and a modem, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0085] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by CPU 301, it performs the various functions defined in the present invention.

[0086] It should be noted that the computer-readable medium shown in the embodiments of the present invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, wherein a computer-readable computer program is carried. The transmitted data signal can take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof.

[0087] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0088] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the system described in the above embodiments; or it may exist independently and not assembled into the system. The storage medium carries one or more computer programs that, when executed by a processor of a system, cause the system to implement the methods provided in the above embodiments.

[0089] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0090] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0091] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0092] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A follow-up timing prediction method based on machine learning, characterized in that, include: Obtain patient information, treatment plan data, and historical follow-up records to obtain the raw medical dataset; Based on the original medical dataset, patient basic information, disease characteristics, treatment response indicators, and physiological parameter change trends are extracted to obtain a standardized feature vector. The standardized feature vector is input into a preset prediction model to calculate the patient's follow-up risk score and determine the optimal follow-up time. Based on the follow-up risk score and the optimal follow-up timing, the follow-up frequency and content are dynamically adjusted to obtain a personalized follow-up plan. Based on the personalized follow-up plan, follow-up tasks are automatically generated at the predicted time and pushed to relevant medical staff to obtain follow-up execution results; Based on the follow-up results, follow-up effect data are collected and the parameters of the prediction model are updated to obtain an optimized prediction model.

2. The method according to claim 1, characterized in that, The step of dynamically adjusting the follow-up frequency and content based on the follow-up risk score and the optimal follow-up timing to obtain a personalized follow-up plan specifically includes: Based on the follow-up risk score, the corresponding follow-up frequency for the patient is determined; Based on the disease characteristics and the treatment response indicators, relevant follow-up items are selected from a preset follow-up content library; Based on the trends in the physiological parameters, key indicators requiring monitoring were identified, and the priority of follow-up content was adjusted accordingly. Based on the prediction results of the optimal follow-up timing, follow-up time points and time windows are generated; The follow-up frequency, follow-up items, key monitoring indicators, and follow-up time are integrated to form a structured personalized follow-up plan.

3. The method according to claim 2, characterized in that, The step of automatically generating follow-up tasks at predicted times and pushing them to relevant medical staff according to the personalized follow-up plan, and obtaining the follow-up execution results, specifically includes: Based on the time nodes in the personalized follow-up plan, a phased follow-up plan containing multiple follow-ups is generated; Configure the triggering parameters for the follow-up task, including the enrollment time, the most recent discharge time, the last follow-up time, and the time period during which the follow-up can be performed; Acquire real-time clinical data from patients and compare it with the monitoring indicators in the follow-up protocol; Calculate the deviation of follow-up data, and generate a higher priority follow-up task when the monitoring indicator exceeds the preset threshold; The follow-up task will be pushed to the work platform of the medical staff corresponding to the patient; The system receives various indicator data collected during the follow-up process as the follow-up execution results, automatically evaluates the follow-up execution results, and generates a follow-up value trend analysis report.

4. The method according to claim 3, characterized in that, The process of generating the personalized follow-up plan also includes: Based on the aforementioned follow-up risk score, assess whether the patient needs to be referred to a higher-level medical institution; When the follow-up risk score exceeds the transfer threshold, the follow-up frequency and follow-up content are automatically adjusted, and preset transfer-related monitoring indicators are added. Calculate the optimal transfer time and target receiving hospital based on the patient's geographical location information and clinical needs; Based on regional medical resource distribution data and real-time traffic information, the optimal transfer route is generated. Based on the optimal transfer timing, the target receiving hospital, the optimal transfer route, the follow-up frequency, and the follow-up content, a personalized follow-up plan including an emergency transfer contingency plan is formed.

5. The method according to claim 3, characterized in that, Before pushing the follow-up task to the patient's corresponding medical staff work platform, the method further includes: Obtain resource allocation information and medical staff scheduling information from medical institutions to form a resource availability database; The follow-up tasks in the aforementioned phased follow-up plan are integrated into a preliminary follow-up task list; Based on the resource availability database and the preliminary follow-up task list, the resource load situation in each time period is calculated to obtain a resource load distribution map; Using the resource load distribution map and the time window in the personalized follow-up plan, the specific execution time of the follow-up task is adjusted to generate an optimized follow-up task plan. Based on the optimized follow-up task plan and the professional matching degree of medical staff, follow-up tasks are assigned to the corresponding medical staff to form the final follow-up task.

6. The method according to claim 5, characterized in that, The step of calculating the resource load for each time period based on the resource availability database and the preliminary follow-up task list to obtain a resource load distribution map specifically includes: Extract the number and professional categories of available medical personnel for each time period from the resource availability database; The number of follow-up tasks in each time period of the preliminary follow-up task list is counted, and the professional category required for each task is determined based on the follow-up content. Based on the professional categories required for each follow-up task, calculate the demand for medical staff in different professional categories for each time period; By comparing the demand for medical staff in each time period with the number of available medical staff in the corresponding time period, the resource gap or surplus of each professional category can be obtained. Based on the aforementioned resource gaps or surpluses, calculate the utilization rate of specialized resources and the overall resource balance for each time period; By arranging the utilization rate of professional resources and the overall resource balance in each time period in chronological order, a resource load distribution map in the time dimension is constructed.

7. The method according to claim 6, characterized in that, The step of adjusting the specific execution time of the follow-up task and generating an optimized follow-up task plan by utilizing the resource load distribution map and the time window in the personalized follow-up plan specifically includes: Identify resource overload periods where resource utilization exceeds a preset threshold and resource idle periods where resource utilization is below a preset threshold from the resource load distribution map; Extract the executable time window and priority score based on clinical needs for each follow-up task from the personalized follow-up plan; Based on the severity of the patient's condition and the degree of abnormality of the monitoring indicators, the urgency score of each follow-up task is calculated, and combined with the priority score, a priority list for task scheduling is determined. According to the priority list, high-priority follow-up tasks will be scheduled during periods with sufficient medical resources within the corresponding executable time window; Identify low-urgency follow-up tasks scheduled during resource overload periods and reschedule them to resource-idle periods, provided that the execution time window is not exceeded. All scheduled follow-up tasks are sorted by execution time to generate an optimized follow-up task plan in a time-series format.

8. A follow-up timing prediction system based on machine learning, characterized in that, The system includes: One or more processors and a memory; the memory is coupled to the one or more processors, the memory being used to store computer program code, the computer program code including computer instructions, the one or more processors invoking the computer instructions to cause the system to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instructions are executed on the system, the system performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the system, the system performs the method as described in any one of claims 1-7.

Images