Anning treatment and protection prognosis prediction system and method fusing objective indexes and functional states

Abstract

The invention relates to the technical field of medical health, and discloses an objective index and functional state fused peaceful treatment and protection prognosis prediction system, which comprises a data acquisition and preprocessing module, a prognosis variable screening and risk layering module, a prediction model construction module, a model evaluation module and a report generation and output module. Through multi-dimensional variable screening and column diagram modeling technologies fusing objective biomarkers and functional states, prognosis influence factors of Anning treatment and care patients can be comprehensively covered, traditional functional subjective scores are concerned, objective indexes such as inflammation and renal functions are also included, risk layering and survival probability mapping are deeply carried out, and the prognosis effect of the Anning treatment and care patients is improved. The prediction result can accurately reflect the real prognosis condition of the peaceful patient, and a more scientific basis is provided for clinical decisions such as immediate care transfer and medical resource allocation.

Description

Technical Field

[0001] This invention relates to the field of medical and health technology, and more specifically discloses a palliative care prognosis prediction system and method that integrates objective indicators and functional status. Background Technology

[0002] With the acceleration of my country's aging population and the increasing demand for end-stage patient care, the importance of palliative care is becoming increasingly prominent.

[0003] The prior art patent document with authorization announcement number CN120260840A discloses "a method for palliative care intervention for hematologic malignancies based on potential profile analysis", which includes the following steps: S1, constructing a multidimensional intervention feature dataset; S2, constructing a profile analysis model based on the feature dataset; S3, performing category stability verification based on the profile analysis results; S4, selecting intervention reference templates based on the stability verification results; S5, matching individual characteristics with intervention reference templates to form a preset intervention path; S6, using the preset intervention path to construct a profile-driven path execution and feedback control system; The patent document with authorization announcement number CN120452736A discloses "a method for predicting the survival of non-tumor palliative care patients based on clinical data". The method includes collecting, organizing and standardizing clinical data of non-tumor palliative care deceased patients during their previous hospitalizations; using the maximum information coefficient (MIC) to select and reduce attributes of high-dimensional data from the perspectives of relevance and redundancy; applying Pearson correlation coefficient and Spearman rank correlation coefficient to perform attribute weighting respectively; and applying a weighted Naive Bayes classification model for modeling.

[0004] While existing technologies can objectively analyze clinical factors related to the survival of non-oncology palliative care patients, predict survival through feature selection using maximum information coefficients and weighted Naive Bayes models, and provide guidance for the formulation of clinical access standards and treatment plans, they can also improve the ability to express profile structures and characterize individual heterogeneity in hematologic malignancy palliative care interventions through cointegrated mapping latent profile modeling, and ensure the continuous effectiveness of intervention pathways by combining profile stability verification and feedback control systems. However, existing technologies have limited applicability, cannot simultaneously cover both cancer and non-cancer end-stage patients, and are difficult to adapt to the diverse disease spectrum of palliative care needs in my country. At the same time, the models rely on complex algorithms, resulting in poor interpretability and high bedside operation thresholds. Moreover, most of them do not achieve deep integration of objective biomarkers (such as inflammation and renal function indicators) and functional status scores, resulting in problems such as subjective assessment bias, single data dimensions, and failure to clearly verify clinical net benefits through decision curve analysis. They cannot meet the individualized needs of short-term survival prediction for palliative care patients, nor can they effectively assist in clinical decision-making and the optimal allocation of medical resources. Summary of the Invention

[0005] The present invention mainly provides a palliative care prognosis prediction system and method that integrates objective indicators and functional status, which can solve the problems raised in the background art.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution, more specifically, a method for predicting the prognosis of palliative care by integrating objective indicators and functional status, comprising: S1. Identify eligible palliative care patients, collect relevant data on these patients, group the collected data, and preprocess the collected data. S2. From all the preprocessed data, key variables related to patient prognosis are selected, and then the selected key variables are classified at the risk level. S3. Based on the selected and segmented key risk variables, build a model to predict the patient's survival probability and present the model in a visual form; S4. The established prediction model is tested from multiple aspects, including discriminative ability, calibration status, and clinical applicability. S5. Based on the model's test results and predictions, generate a prognostic prediction report for each individual patient and store the patient's relevant data and prediction results for easy follow-up.

[0007] Furthermore, in S1, eligible palliative care patients must meet preset inclusion and exclusion criteria; the collected data includes 30 candidate variable data and overall survival follow-up data; the grouping is carried out by random grouping in a 7:3 ratio to divide the training set and the validation set; the preprocessing includes data integrity verification and baseline feature balance verification of the two groups.

[0008] Furthermore, in S2, the key variable screening adopts the LASSO Cox regression algorithm, and the key variables screened are C-reactive protein, estimated glomerular filtration rate, and Bartholomew's daily living ability score; the risk level division adopts the maximum selection rank statistic to divide the key variables into high-risk groups and low-risk groups.

[0009] Furthermore, in S4, the discriminant ability test uses a time-dependent ROC curve, the calibration test uses a calibration curve, and the clinical applicability test uses decision curve analysis.

[0010] According to another aspect of the present invention, a palliative care prognosis prediction system integrating objective indicators and functional status is provided. This system is implemented based on the above-mentioned palliative care prognosis prediction method integrating objective indicators and functional status, and specifically includes: a data acquisition and preprocessing module, a prognostic variable screening and risk stratification module, a prediction model construction module, a model evaluation module, and a report generation and output module. The data acquisition and preprocessing module identifies eligible palliative care patients and collects relevant data, performing grouping and preprocessing operations on the data. The prognostic variable screening and risk stratification module filters key variables related to patient prognosis from the preprocessed data and classifies the key variables into risk levels. The prediction model construction module builds a model for predicting patient survival probability based on the classified key variables and displays the model in a visual form. The model evaluation module performs multi-faceted testing on the built prediction model from the dimensions of discriminative ability, calibration, and clinical applicability. The report generation and output module generates a prognosis prediction report for a single patient based on the model's testing results and prediction performance, and stores the patient-related data and prediction results for subsequent traceability.

[0011] Furthermore, the data acquisition and preprocessing module includes: a patient enrollment and screening module, a data acquisition module, and a data preprocessing module; Patient enrollment and screening module: Screens eligible palliative care patients based on preset inclusion and exclusion criteria; Data acquisition module: Collects data on 30 candidate variables and overall survival follow-up data from eligible patients; Data preprocessing module: Randomly groups the collected data in a 7:3 ratio and performs data integrity verification and baseline feature balance verification between the two groups.

[0012] Furthermore, the prognostic variable screening and risk stratification module includes: a LASSO Cox regression screening module and a risk stratification module; LASSO Cox Regression Screening Module: Uses the LASSO Cox regression algorithm to screen out key variables related to patient prognosis from preprocessed data; Risk stratification module: The maximum selection rank statistic is used to divide the selected key variables into high-risk and low-risk groups.

[0013] Furthermore, the prediction model building module includes: a multi-factor Cox regression modeling module, a nomogram visualization module, and a survival probability prediction module; Multivariate Cox regression modeling module: Based on the well-defined key variables of risk, a multivariate Cox regression algorithm is used to build a model to predict the survival probability of patients; Nodal plot visualization module: Visualizes the completed patient survival probability prediction model in the form of a nodal plot; Survival probability prediction module: Based on a visualized nomogram model, it calculates and outputs the 30-day, 60-day, and 90-day survival probabilities of patients.

[0014] Furthermore, the model evaluation module includes: a discriminative power evaluation module, a calibration evaluation module, and a clinical usability evaluation module; Discriminant power evaluation module: The discriminant power of the constructed prediction model is tested using time-dependent ROC curves; Calibration evaluation module: The calibration curve is used to verify the calibration status of the built prediction model; Clinical applicability assessment module: Decision curve analysis is used to test the clinical applicability of the established predictive model.

[0015] Furthermore, the report generation and output module includes: a personalized prediction report generation module, a decision suggestion module, and a data storage and traceability module; Personalized prediction report generation module: Based on model validation results and patient survival probability predictions, it generates a prognostic prediction report for each individual patient. Decision suggestion module: Provides clinical decision-making suggestions based on patient risk stratification and survival probability prediction results; Data storage and traceability module: Stores relevant patient data and model prediction results for subsequent traceability and retrieval.

[0016] The beneficial effects of this invention's palliative care prognosis prediction system and method, which integrates objective indicators and functional status, are as follows: By integrating multi-dimensional variable screening and nomogram modeling techniques of objective biomarkers and functional status, it can comprehensively cover the prognostic influencing factors of palliative care patients. It not only focuses on traditional subjective functional scores but also incorporates objective indicators such as inflammation and renal function, and conducts in-depth risk stratification and survival probability mapping, making the prediction results more accurately reflect the true prognosis of palliative care patients. This provides a more scientific basis for clinical decisions such as end-of-life care referrals and medical resource allocation. Furthermore, through LASSO... The variable screening and stratification technique combining Cox regression and maximum selection rank statistic achieves precise screening of candidate variables and effective classification of high / low risk, breaking the limitations of variable redundancy and subjective stratification. This significantly improves the prognostic correlation and stratification discrimination of model variables, enhancing the accuracy and specificity of prognostic predictions and increasing the model's universality across different disease spectrums. Furthermore, the clinical application technique combining nomogram visualization and decision curve analysis provides a more transparent and user-friendly predictive tool, clearly demonstrating the correlation between the risk of various patient indicators and the probability of survival. Additionally, decision support based on clinical net benefit validation can rapidly provide scientific and effective intervention recommendations in palliative care scenarios, greatly improving the decision-making level of palliative care and the quality of patient prognosis management. Attached Figure Description

[0017] The present invention will now be described in further detail with reference to the accompanying drawings and specific implementation methods.

[0018] Figure 1 This is a schematic diagram of the system framework; Figure 2 This is a flowchart illustrating the method. Figure 3 This is a schematic diagram of the coefficient path; Figure 4 This is a diagram illustrating cross-validation. Figure 5 A schematic diagram illustrating the optimal cutoff value analysis for CRP; Figure 6 A schematic diagram illustrating the analysis of the optimal cutoff value for eGFR; Figure 7 A schematic diagram illustrating the optimal cutoff value analysis for BADL; Figure 8 A schematic diagram of Kaplan-Meier survival curves for CRP stratification; Figure 9 A schematic diagram of Kaplan-Meier survival curves for eGFR stratification; Figure 10 A schematic diagram of the stratified Kaplan-Meier survival curves for BADL; Figure 11A nomogram for predicting the 30-day, 60-day, and 90-day survival probabilities of patients receiving palliative care; Figure 12 A schematic diagram of the timeROC curve for the training set; Figure 13 A schematic diagram of the timeROC curve for the validation set; Figure 14 This is a schematic diagram of the 30-day, 60-day, and 90-day DCA curves for the prediction model. Detailed Implementation

[0019] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0020] According to one aspect of the invention, such as Figures 1-14 As shown, a system and method for predicting the prognosis of palliative care that integrates objective indicators and functional status are provided, including: Step 1: Data Acquisition and Preprocessing Identify eligible palliative care patients, collect relevant data on these patients, group the collected data, and preprocess the collected data. Specifically, eligible palliative care patients must meet the pre-defined inclusion and exclusion criteria; the collected data includes data on 30 candidate variables and overall survival follow-up data; the grouping is done by randomization in a 7:3 ratio to divide the training set and the validation set; the preprocessing includes data integrity verification and baseline feature balance verification between the two groups.

[0021] The study participants were precisely screened based on the pre-set inclusion and exclusion criteria. The inclusion criteria were patients diagnosed with irreversible end-stage diseases (such as advanced cancer, chronic organ failure, etc.), with an expected survival of ≤6 months at the time of enrollment, and the patient and their legal representative signed a palliative care consent form (clearly giving up curative treatment). The exclusion criteria included patients whose condition improved unexpectedly, effective treatments became available, patients who wished to withdraw palliative care, patients with missing key information in their baseline data, or patients who were lost to follow-up within 30 days after enrollment. In addition, multi-dimensional information collection was carried out, specifically covering 30 candidate variables and outcome variables. Candidate variables included demographic characteristics (age, sex, BMI, etc.), clinical status (pain level, cognitive function, etc.), objective laboratory indicators (C-reactive protein, estimated glomerular filtration rate, etc.), and functional assessment (Bartholin's score of activities of daily living, etc.). The outcome variable was overall survival, defined as "the time from the first time a patient receives palliative care to death from any cause." For patients lost to follow-up, the last follow-up date was used as the censoring time. The timing of data collection was clearly defined as baseline data being directly extracted from the electronic medical record system on the day the patient received palliative care. Laboratory indicators were the results of routine tests completed within 24 hours of admission. Overall survival was tracked and obtained through monthly telephone follow-ups or outpatient follow-ups until the pre-specified study endpoint. Meanwhile, a random grouping algorithm was used to divide the collected qualified data into a training set (for the development and construction of subsequent prediction models) and a validation set (for external validity verification of the models) in a 7:3 ratio. Finally, the data quality control process is executed. First, data integrity is verified, and data entries with serious missing or abnormalities are removed. Then, statistical methods such as t-test, Mann-Whitney U test, and chi-square test are used to verify whether the baseline characteristics (such as age distribution, gender ratio, disease type, survival status, etc.) of the training set and validation set are balanced and comparable. Finally, Kaplan-Meier survival curves are used to verify that there is no significant difference in the overall survival of the two groups of patients (as shown in Table 1 below). Table 1 Baseline characteristics of patients receiving palliative care Ensure the scientific validity and reliability of data grouping to provide standardized data input for model development.

[0022] Step 2: Key Variable Screening and Risk Stratification From all the preprocessed data, key variables related to patient prognosis were screened out, and then the screened key variables were classified at the risk level. Specifically, the LASSO Cox regression algorithm was used to screen key variables, and the key variables selected were C-reactive protein, estimated glomerular filtration rate, and Bartholomew's daily living ability score; the maximum selection rank statistic was used to divide the key variables into high-risk and low-risk groups.

[0023] The prognostic variable screening and risk stratification module screens key variables on the preprocessed training set data, using the LASSO Cox regression algorithm to analyze 30 candidate variables. The algorithm is implemented using the glmnet package in R language, and 10-fold cross-validation is performed based on survival outcomes (see attached). Figure 3Appendix Figure 4 (as shown) During the cross-validation process, the optimal regularization parameter λ was determined based on the "1 standard error rule". The coefficients of the variables were compressed using this parameter, and the coefficients of variables without prognostic association were zeroed out. Finally, three key variables with independent prognostic value were selected: C-reactive protein (CRP), estimated glomerular filtration rate (eGFR), and Bartholomew's Daily Living Skills (BADL) score. In addition, after variable screening, risk stratification was further carried out. The maximum selection rank statistic was used to calculate the optimal prognostic cutoff value for the three selected continuous key variables. Statistical analysis was performed using the maxstat package in R language to determine the critical values ​​for distinguishing between high and low mortality risk for each variable (as shown in the appendix). Figure 5 Appendix Figure 6 Appendix Figure 7 (as shown) Finally, based on the calculated cutoff values, each key variable was divided into high-risk and low-risk groups, forming a standardized risk stratification result. The survival differences between the high- and low-risk groups for each variable can be observed using Kaplan-Meier survival curves (see attached). Figure 8 Appendix Figure 9 Appendix Figure 10 As shown in the figure, this provides a clear basis for variable inputs for the construction of subsequent prediction models.

[0024] Step 3: Construction and Visualization of Survival Probability Prediction Model Based on the screening and classification of key risk variables, a model for predicting patient survival probability is built, and the model is presented in a visual form. Specifically, the prediction model building module uses the high / low risk stratification variables obtained in step 2 as the core input, and uses a multivariate Cox regression algorithm to build a basic prediction model. It then performs correlation analysis between the risk stratification status of C-reactive protein, estimated glomerular filtration rate, and Bartholomew's daily living ability score and the patient's survival outcome.

[0025] The modeling process uses only the training set data to run the algorithm, outputs the hazard ratio (HR) of each key variable through regression analysis, and determines the weight of each variable in the model based on the HR value, which serves as the core basis for the subsequent scoring rules. Meanwhile, the nomogram visualization module transforms the completed regression model into a graphical tool, designing a hierarchical structure of "indicator-score-total score-survival probability," with clear score scales marked on the corresponding axis for each key variable's high / low risk group (as shown in the attached figure). Figure 11 (as shown) Furthermore, by establishing a pre-defined mapping relationship between the total score and the model's linear prediction value, the patient's cumulative score is directly converted into a specific survival probability, covering three short-term prognostic time points: 30 days, 60 days, and 90 days, thus achieving precise quantification of survival risk at different time periods.

[0026] Step 4: Multi-dimensional testing of the prediction model The established predictive model was tested from multiple perspectives, including its discriminative ability, calibration status, and clinical applicability. Specifically, the discriminant ability test uses a time-dependent ROC curve, the calibration test uses a calibration curve, and the clinical applicability test uses a decision curve analysis.

[0027] The discriminant evaluation module conducts tests on both the training and validation sets. The AUC values ​​of the model at three time points (30 days, 60 days, and 90 days) are calculated using time-dependent ROC curves. The AUC values ​​on the training set are 0.764, 0.716, and 0.705, respectively, and the AUC values ​​on the validation set are 0.770, 0.748, and 0.788, respectively, all greater than 0.7, demonstrating that the model possesses good risk discrimination ability (see attached). Figure 12 Appendix Figure 13 (as shown) Meanwhile, the calibration evaluation module compares the survival probability predicted by the model with the actual observed survival probability by plotting a calibration curve. The data points in both the training and validation sets are closely distributed around the ideal 45-degree reference line, indicating that the predicted values ​​are highly consistent with the actual outcomes and the model calibration effect is excellent. Furthermore, the clinical applicability assessment module employed decision curve analysis (DCA) to calculate net benefit at different clinical risk thresholds. Results showed that within the 20%-40% risk threshold range for 90-day survival prediction, the model's net benefit was significantly higher than the "full intervention" or "no intervention" strategies, confirming the model's clear practical value within commonly used clinical decision intervals (e.g., ...). Figure 14 (as shown) Finally, combining the test results of the two datasets, the model showed stability in terms of discriminative power, calibration degree, and clinical applicability, with no significant differences. This further proves that the model can be extended to different palliative care patient groups, providing a reliable basis for clinical application.

[0028] Step 5: Generation and Data Storage of Prognostic Prediction Reports Based on the model's test results and predictions, a prognostic prediction report for each individual patient is generated, and the patient's relevant data and prediction results are stored for easy follow-up. Specifically, based on the results of model testing, combined with the risk stratification of key variables and the predicted survival probability of an individual patient, a personalized prognostic report is generated, and the patient's full-process data is stored simultaneously for subsequent query and traceability.

[0029] The personalized prediction report generation module integrates the patient's core information, including the specific values ​​of C-reactive protein (CRP), estimated glomerular filtration rate (eGFR), Bartholomew's Daily Living Skills (BADL) score and corresponding risk stratification (high / low risk), the total risk score calculated by the model, and the specific survival probabilities at 30 days, 60 days, and 90 days. It also briefly marks the key indicators of model testing (such as the validation set AUC value) to allow medical staff to clearly understand the basis and reliability of the prediction. The decision-making recommendation module provides targeted clinical recommendations based on the patient's risk stratification and survival probability. For example, for high-risk patients with a 90-day survival probability of less than 20%, it is recommended to prioritize the initiation of the end-of-life care referral process. For patients with a survival probability of more than 40%, it is recommended to focus on optimizing symptom relief plans such as pain control and nutritional support, and assist doctors and patients in jointly developing a care plan. In addition, the data storage and traceability module will store the patient's full-process data in a structured manner, covering the original data of 30 candidate variables at the baseline stage, the screening results of key variables, risk stratification records, model prediction results, and the actual survival outcomes obtained in subsequent follow-ups. All data are associated with the patient's unique identifier (such as electronic medical record ID) and support multi-dimensional retrieval by patient ID, enrollment time, disease type, etc. Meanwhile, the stored data will be archived in a standardized format, which not only meets the needs of clinical retrospective analysis, but also provides real-world data support for subsequent model iterations. For example, after accumulating a certain number of new cases, the model parameters can be re-validated or fine-tuned based on the stored data to further improve the accuracy of predictions.

[0030] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention are also within the protection scope of the present invention.

Claims

1. A method for predicting the prognosis of palliative care by integrating objective indicators and functional status, characterized in that, The method includes: S1. Identify eligible palliative care patients, collect relevant data on these patients, group the collected data, and preprocess the collected data. S2. From all the preprocessed data, key variables related to patient prognosis are selected, and then the selected key variables are classified at the risk level. S3. Based on the selected and segmented key risk variables, build a model to predict the patient's survival probability and present the model in a visual form; S4. The established prediction model is tested from multiple aspects, including discriminative ability, calibration status, and clinical applicability. S5. Based on the model's test results and predictions, generate a prognostic prediction report for each individual patient and store the patient's relevant data and prediction results for easy follow-up.

2. The method for predicting the prognosis of palliative care by integrating objective indicators and functional status according to claim 1, characterized in that: Patients eligible for palliative care in S1 must meet the preset inclusion and exclusion criteria; the collected data includes data on 30 candidate variables and overall survival follow-up data. The data was randomly grouped into training and validation sets in a 7:3 ratio; preprocessing included data integrity verification and baseline feature balance verification between the two sets.

3. The method for predicting the prognosis of palliative care by integrating objective indicators and functional status according to claim 1, characterized in that: The key variables in S2 were selected using the LASSO Cox regression algorithm. The key variables selected were C-reactive protein, estimated glomerular filtration rate, and Bartholomew's daily living ability score. The risk level was divided into high-risk and low-risk groups using the maximum selection rank statistic.

4. The method for predicting the prognosis of palliative care by integrating objective indicators and functional status according to claim 1, characterized in that: In S4, the discriminant ability test uses a time-dependent ROC curve, the calibration test uses a calibration curve, and the clinical applicability test uses decision curve analysis.

5. A palliative care prognosis prediction system integrating objective indicators and functional status, characterized in that, This system is based on the palliative care prognosis prediction method integrating objective indicators and functional status as described in any one of claims 1-4. Specifically, it includes: a data acquisition and preprocessing module, a prognostic variable screening and risk stratification module, a prediction model construction module, a model evaluation module, and a report generation and output module. The data acquisition and preprocessing module identifies eligible palliative care patients and collects relevant data, performing grouping and preprocessing operations on the data. The prognostic variable screening and risk stratification module filters key variables related to patient prognosis from the preprocessed data and classifies these key variables into risk levels. The prediction model construction module builds a model to predict patient survival probability based on the classified key variables and displays the model visually. The model evaluation module performs multi-faceted testing on the constructed prediction model from the dimensions of discriminative ability, calibration, and clinical applicability. The report generation and output module generates a prognostic prediction report for a single patient based on the model's testing results and prediction performance, and stores the patient-related data and prediction results for subsequent traceability.

6. The palliative care prognosis prediction system integrating objective indicators and functional status according to claim 5, characterized in that: The data acquisition and preprocessing module includes: a patient enrollment and screening module, a data acquisition module, and a data preprocessing module; Patient enrollment and screening module: Screens eligible palliative care patients based on preset inclusion and exclusion criteria; Data acquisition module: Collects data on 30 candidate variables and overall survival follow-up data from eligible patients; Data preprocessing module: Randomly groups the collected data in a 7:3 ratio and performs data integrity verification and baseline feature balance verification between the two groups.

7. The palliative care prognosis prediction system integrating objective indicators and functional status according to claim 5, characterized in that: The prognostic variable screening and risk stratification module includes: a LASSO Cox regression screening module and a risk stratification module; LASSO Cox Regression Screening Module: Uses the LASSO Cox regression algorithm to screen out key variables related to patient prognosis from preprocessed data; Risk stratification module: The maximum selection rank statistic is used to divide the selected key variables into high-risk and low-risk groups.

8. The palliative care prognosis prediction system integrating objective indicators and functional status according to claim 5, characterized in that: The prediction model building module includes: a multi-factor Cox regression modeling module, a nomogram visualization module, and a survival probability prediction module; Multivariate Cox regression modeling module: Based on the well-defined key variables of risk, a multivariate Cox regression algorithm is used to build a model to predict the survival probability of patients; Nodal plot visualization module: Visualizes the completed patient survival probability prediction model in the form of a nodal plot; Survival probability prediction module: Based on a visualized nomogram model, it calculates and outputs the 30-day, 60-day, and 90-day survival probabilities of patients.

9. The palliative care prognosis prediction system integrating objective indicators and functional status according to claim 5, characterized in that: The model evaluation module includes: a discriminant evaluation module, a calibration evaluation module, and a clinical usability evaluation module; Discriminant power evaluation module: The discriminant power of the constructed prediction model is tested using time-dependent ROC curves; Calibration evaluation module: The calibration curve is used to verify the calibration status of the built prediction model; Clinical applicability assessment module: Decision curve analysis is used to test the clinical applicability of the established predictive model.

10. The palliative care prognosis prediction system integrating objective indicators and functional status according to claim 5, characterized in that: The report generation and output module includes: a personalized prediction report generation module, a decision suggestion module, and a data storage and traceability module; Personalized prediction report generation module: Based on model validation results and patient survival probability predictions, it generates a prognostic prediction report for each individual patient. Decision suggestion module: Provides clinical decision-making suggestions based on patient risk stratification and survival probability prediction results; Data storage and traceability module: Stores relevant patient data and model prediction results for subsequent traceability and retrieval.

Images