Establishment method and application of prognosis prediction model for thrombosis high-risk group primary thrombocythemia patients

Abstract

The invention relates to a method for establishing a prognosis prediction model for primary thrombocythemia patients in a thrombus high-risk group and application of the method, establishes a dynamic prognosis line chart for the primary thrombocythemia patients with previous thrombosis history, and aims to provide a more accurate risk assessment tool for clinical decision making. Strict statistical analysis verifies that the dynamic column chart developed by the method shows high prediction accuracy of prognosis of patients with primary thrombocythemia in a thrombus high-risk group, and powerful support is provided for risk assessment and individualized treatment in clinical practice.

Description

Technical Field

[0001] The present invention relates to the field of biomedicine, and in particular to a method for establishing a prognosis prediction model for essential thrombocythaemia patients in a high-risk group of thrombosis and its application. Background Art

[0002] Essential thrombocythemia is a subtype of Philadelphia chromosome-negative myeloproliferative neoplasm characterized by clonal proliferation of the megakaryocyte lineage, resulting in elevated platelet counts. The International Prognostic Scoring System for Thrombosis in Essential Thrombocythemia (IPSET-thrombosis, revised) is a scoring system used to predict the risk of thrombosis in patients with essential thrombocythemia (ET). The model was established based on a multicenter, large-scale clinical data analysis, with the aim of predicting the risk of arterial and venous thrombosis in ET patients by integrating multiple independent risk factors.

[0003] The process of establishing the model can be divided into the following key steps: 1. Data collection and patient selection: The researchers first collected a large number of clinical data of ET patients from multiple centers, including demographic characteristics, laboratory test results, treatment regimens, and the occurrence of thrombotic events. The original data sources include large-scale prospective and retrospective cohort studies. Through strict standard screening, patients who did not meet the diagnostic criteria or were missing key data were excluded. 2. Determine predictive variables: Based on previous studies and clinical experience, the researchers selected factors known to be associated with the risk of thrombosis in ET patients, including whether the age was greater than 60 years old, whether the patient had a history of arterial or venous thrombosis, whether the patient carried the JAK2V617F mutation, and cardiovascular risk factors. 3. Statistical analysis and risk stratification: In the process of establishing the model, the researchers used multivariate Cox regression model analysis to evaluate the independent association between each predictor variable and the occurrence of thrombotic events. Each variable was assigned a different weight or score according to its impact on the risk of thrombosis. According to the analysis results, the revised version of IPSET-thrombosis divides patients into the following three risk levels: low risk: no JAK2 mutation, age ≤ 60 years old, no history of thrombosis; medium risk: age > 60 years old or with JAK2 mutation, but no history of thrombosis; high risk: with a history of thrombosis. 4. Validation and optimization of the model: The revised model evaluates its predictive ability through internal and external validation. During the validation process, the researchers used Kaplan-Meier survival analysis and Cox proportional hazard model to test the discrimination and calibration of the model in different risk groups. In addition, the study also conducted ROC curve analysis to calculate the predictive power of the model (such as AUC value). The revised model showed good predictive ability in multiple validations, and the original model was improved to make it more concise and effective. 5. Clinical application: Finally, IPSET-thrombosis (revised version) was promoted and applied in clinical practice to help doctors formulate personalized prevention and treatment strategies according to the patient's risk level, especially in the choice of antithrombotic therapy.

[0004] Among them, patients with a history of thrombosis account for about 20% of all essential thrombocythemia patients, and this group of people is defined as a high-risk group for thrombosis by IPSET-thrombosis (revised version). At present, the clinical characteristics and prognostic factors of this patient group have yet to be clarified. Therefore, if a method for predicting the prognosis probability of essential thrombocythemia patients in the high-risk group for thrombosis can be established based on cohort data of clinical and laboratory characteristics, it will provide an effective means for clinical precision treatment.

[0005] Previous studies have proposed risk factors for survival or leukemic transformation in patients with essential thrombocythemia, including elevated white blood cell counts, anemia, and splenomegaly. However, there are no studies specifically focusing on prognostic risk factors in patients with essential thrombocythemia who are at high risk of thrombosis and have been followed up for years or decades. In this regard, a comprehensive assessment of risk factors is crucial for the clinical management of each patient and is of great significance for the development of precise treatment plans, which is one of the urgent issues to be addressed in this field. Summary of the invention

[0006] To solve the above technical problems, the present invention provides a method for establishing a prognostic prediction model for essential thrombocythemia patients in a high-risk group for thrombosis and its application. A prognostic prediction model is established by this method, and a dynamic nomogram is generated. The dynamic nomogram comprehensively considers multiple key prognostic factors, and has high accuracy in predicting the prognosis of essential thrombocythemia patients in a high-risk group for thrombosis, providing strong support for risk assessment and individualized treatment in clinical practice.

[0007] To achieve this object, the present invention adopts the following technical solutions:

[0008] In a first aspect, the present invention provides a method for establishing a prognosis prediction model for essential thrombocythaemia patients in a high-risk group for thrombosis, the method comprising the following steps:

[0009] (1) Included medical records of patients with essential thrombocythemia who had a history of thrombosis;

[0010] (2) Follow up the patients for a long time and collect clinical data during the follow-up period, including survival time, recurrence or death;

[0011] (3) Use multiple imputation to handle missing values ​​in the data;

[0012] (4) Screen out potential candidate variables related to disease prognosis from existing data based on literature and experience, and standardize or normalize numerical variables;

[0013] (5) Perform univariate Cox regression analysis on each candidate variable to preliminarily evaluate the relationship between each variable and survival time or prognosis, obtain the regression coefficient and significance level. The regression coefficient represents the effect of the variable on the risk of event occurrence. Variables with a significance level P value less than 0.05 are selected as candidate variables for further multivariate regression analysis.

[0014] (6) The significant variables screened out in the univariate Cox regression analysis were included in the multivariate Cox regression model to evaluate the independent prognostic effect of these variables after controlling other factors, and the regression coefficient and significance level were obtained. The regression coefficient represents the influence of each variable, and the variables with a significance level P value less than 0.05 were selected;

[0015] (7) Based on the Cox regression results, the DynNom algorithm was used to generate a dynamic nomogram.

[0016] Preferably, the multiple interpolation method in step (3) is a multiple interpolation method of chain equations.

[0017] Preferably, the missing proportion imputed by the multiple imputation method in step (3) is ≤ 20% (for example, it may be 1%, 5%, 10%, 15% or 20%, etc.).

[0018] Preferably, the candidate variables in step (4) include any one or a combination of at least two of age, gender, blood cell count, white blood cell count, hemoglobin count, biochemical parameters, spleen size by abdominal color Doppler ultrasound, bone marrow hematopoietic cell chromosome karyotype or JAK2 mutation.

[0019] Preferably, the variables with a significance level P value less than 0.05 in step (6) include whether the age is less than 60 years old, whether the spleen is enlarged, whether the chromosome karyotype is normal, any one of the white blood cell count and hemoglobin count, or a combination of at least two of them.

[0020] Preferably, step (6) further comprises adjusting the model using AIC / BIC.

[0021] In a second aspect, the present invention provides a prognosis prediction model for essential thrombocythaemia patients in a high-risk group for thrombosis, wherein the prognosis prediction model is constructed using the establishment method described in the first aspect.

[0022] Preferably, the prognosis prediction model comprises:

[0023] Data collection module: used to collect case data of patients with essential thrombocythemia who have a history of thrombosis, conduct long-term follow-up of the patients, and collect clinical data during the follow-up period, including survival time, recurrence or death;

[0024] Data processing module: used to handle missing values ​​in data using multiple imputation methods;

[0025] Variable screening and processing module: used to screen out potential candidate variables related to disease prognosis from existing data based on literature and experience, and to standardize or normalize numerical variables;

[0026] Univariate Cox regression analysis module: used to perform univariate Cox regression analysis on each candidate variable, preliminarily evaluate the relationship between each variable and survival time or prognosis, obtain regression coefficients and significance levels. The regression coefficients represent the impact of the variables on the risk of events. Variables with a significance level P value less than 0.05 are selected as candidate variables for further multivariate regression analysis.

[0027] Multivariate Cox regression analysis module: used to incorporate the significant variables screened out in the univariate Cox regression analysis into the multivariate Cox regression model, evaluate the independent prognostic effects of these variables after controlling other factors, obtain the regression coefficient and significance level. The regression coefficient represents the influence of each variable, and select variables with a significance level P value less than 0.05;

[0028] Model output module: used to generate dynamic nomograms based on Cox regression results using the DynNom algorithm.

[0029] Preferably, the multiple interpolation method is a multiple interpolation method of chain equations.

[0030] Preferably, the multiple imputation method interpolates missing proportions ≤ 20% (eg, 1%, 5%, 10%, 15% or 20%, etc.).

[0031] Preferably, the candidate variables in the variable screening and processing module include any one or a combination of at least two of age, gender, blood cell count, white blood cell count, hemoglobin count, biochemical parameters, spleen size by abdominal color Doppler ultrasound, bone marrow hematopoietic cell chromosome karyotype or JAK2 mutation.

[0032] Preferably, the variables with a significance level P value less than 0.05 in the multi-factor Cox regression analysis module include whether the age is less than 60 years old, whether the spleen is enlarged, whether the chromosome karyotype is normal, any one of the white blood cell count and the hemoglobin count, or a combination of at least two of them.

[0033] Preferably, the multi-factor Cox regression analysis module also includes using an AIC / BIC adjustment model.

[0034] In a third aspect, the present invention provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implements the steps in the establishment method described in the first aspect.

[0035] In a fourth aspect, the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program or instructions, and when the computer program or instructions are executed by the processor, the steps in the establishment method described in the first aspect are implemented.

[0036] Other specific point values ​​within the above numerical ranges can be selected and will not be described in detail here.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The present invention has developed a dynamic prognostic nomogram for patients with essential thrombocythemia in the high-risk group for thrombosis, aiming to provide a more accurate risk assessment tool for clinical decision-making. The dynamic nomogram comprehensively considers multiple key prognostic factors, including age (divided into two groups of <60 years and ≥60 years), the presence or absence of splenomegaly, elevated white blood cell counts, decreased hemoglobin levels, and chromosome karyotypes (abnormal or normal), in order to achieve a multi-dimensional assessment of the prognosis of patients with essential thrombocythemia in the high-risk group for thrombosis. Through rigorous statistical analysis and verification, the dynamic nomogram developed by the present invention shows a high predictive accuracy for the prognosis of patients with essential thrombocythemia in the high-risk group for thrombosis, providing strong support for risk assessment and individualized treatment in clinical practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 The present invention is a flowchart of the method for establishing the prognosis prediction model.

[0040] Figure 2 This is a graph showing the evaluation results of the 3-year, 5-year and 10-year survival rate prediction accuracy of the prognostic prediction model of the present invention.

[0041] Figure 3 This is a diagram of the internal validation results of the prognosis prediction model of the present invention. DETAILED DESCRIPTION

[0042] To further illustrate the technical means and effects of the present invention, the present invention is further described below in conjunction with the embodiments and drawings. It should be understood that the specific implementation methods described herein are only used to explain the present invention, rather than to limit the present invention.

[0043] If no specific techniques or conditions are specified in the examples, the techniques or conditions described in the literature in the field or the product instructions are used. If no manufacturer is specified for the reagents or instruments used, they are all conventional products that can be purchased through regular channels.

[0044] Example 1

[0045] This embodiment provides a method for establishing a prognostic prediction model for essential thrombocythemia patients in a high-risk group for thrombosis. The establishment process is as follows: Figure 1 As shown, the establishment method comprises the following steps:

[0046] 1. Inclusion of eligible clinical case data: First, it is necessary to identify patient cases that meet the research conditions and ensure the completeness and applicability of the data. These conditions include the diagnostic criteria for essential thrombocythemia, specific follow-up time, and previous history of thrombosis.

[0047] 2. Follow-up: Long-term follow-up of patients and collection of clinical data during the follow-up period, which usually include survival time, recurrence, death and other related endpoint events.

[0048] 3. Multiple imputation: To deal with missing values ​​in the data, which are common in follow-up data, multiple imputation can reduce the bias caused by missing values. The imputation missing ratio needs to be ≤20%. Exceeding 20% ​​may affect the reliability of the model results. The imputation method is Multiple Imputation by Chained Equations (MICE).

[0049] 4. Candidate clinical characteristics and variables: Potential candidate variables were screened from existing data, including demographic information (such as age and gender), laboratory test results (such as blood routine, organ function, etc.) and gene mutation status. The candidate variables were selected based on literature and experience to select variables related to disease prognosis, including age, gender, blood cell count, biochemical parameters, abdominal ultrasound spleen size, bone marrow hematopoietic cell chromosome karyotype, JAK2 mutation, etc. At the same time, numerical variables were standardized or normalized.

[0050] 5. Univariate Cox regression analysis: Univariate Cox regression analysis was performed on each candidate variable to preliminarily evaluate the relationship between each variable and survival time or prognosis. The regression coefficient β and significance level P were obtained. The regression coefficient represents the effect of the variable on the risk of event occurrence. P<0.05 was usually selected as the significance threshold. Variables with a P value less than 0.05 in the univariate analysis were selected as candidate variables for further multivariate regression analysis.

[0051] 6. Multivariate Cox regression analysis: The significant variables screened out in the univariate analysis were included in the multivariate Cox regression model to evaluate the independent prognostic effect of these variables after controlling other factors. The regression coefficient β and significance level P were obtained. The regression coefficient represents the influence of each variable. P<0.05 is usually selected to indicate significance. Variables with a significance level P value less than 0.05 were selected, and finally the model was adjusted by AIC / BIC or other model selection methods.

[0052] 7. DynNom algorithm: Based on the Cox regression results, the DynNom algorithm is used to generate a dynamic nomogram to help predict individualized survival probability. The dynamic nomogram can dynamically generate prediction results based on the patient's characteristic values.

[0053] 8. Online calculator for survival prediction model: Based on the constructed nomogram, using the developed online tool (https: / / cae29l-jia-chen.shinyapps.io / DynNomapp / ), doctors or researchers can input the data of specific patients and generate individualized survival predictions. By inputting all the significant variables included in the model, you can see the survival curve generated based on the characteristics of the input patient parameters, as well as the individual survival probability prediction at different follow-up time points, such as the prediction of 3-year, 5-year, and 10-year survival probability.

[0054] Example 2

[0055] This example uses the establishment method provided in Example 1 to obtain a prognosis prediction model, and the prognosis prediction model includes:

[0056] Data collection module: used to collect case data of patients with essential thrombocythemia who have a history of thrombosis, conduct long-term follow-up of the patients, and collect clinical data during the follow-up period, including survival time, recurrence or death;

[0057] Data processing module: used to process missing values ​​in data using multiple imputation method of chain equations, with the imputation missing ratio ≤ 20%;

[0058] Variable screening and processing module: used to screen out potential candidate variables related to disease prognosis from existing data based on literature and experience. Candidate variables include age, gender, blood cell count, white blood cell count, hemoglobin count, biochemical parameters, spleen size by abdominal color Doppler ultrasound, bone marrow hematopoietic cell chromosome karyotype or JAK2 mutation, any one or a combination of at least two of them, and perform standardization or normalization on numerical variables;

[0059] Univariate Cox regression analysis module: used to perform univariate Cox regression analysis on each candidate variable, preliminarily evaluate the relationship between each variable and survival time or prognosis, obtain regression coefficients and significance levels. The regression coefficients represent the impact of the variables on the risk of events. Variables with a significance level P value less than 0.05 are selected as candidate variables for further multivariate regression analysis.

[0060] Multivariate Cox regression analysis module: used to incorporate the significant variables screened out in the univariate Cox regression analysis into the multivariate Cox regression model, evaluate the independent prognostic effects of these variables after controlling other factors, obtain the regression coefficient and significance level. The regression coefficient represents the influence of each variable. Variables with a significance level P value less than 0.05 are selected, including whether the age is less than 60 years old, whether the spleen is enlarged, whether the chromosome karyotype is normal, any one of the white blood cell count and hemoglobin count, or a combination of at least two of them, and the AIC / BIC is used to adjust the model;

[0061] Model output module: used to generate dynamic nomograms based on Cox regression results using the DynNom algorithm.

[0062] Test Example 1

[0063] This test example evaluates the accuracy of the 3-year, 5-year, and 10-year survival rate predictions based on the prognosis prediction model provided in Example 2. The results are as follows: Figure 2 As shown in the figure, the AUC values ​​and their 95% confidence intervals of the model's 3-year, 5-year, and 10-year survival rate prediction accuracy are 0.854 (95% CI: 0.681-1.000), 0.769 (95% CI: 0.632-0.906), and 0.813 (95% CI: 0.681-0.944), respectively. The specific steps are as follows:

[0064] (1) Data preparation:

[0065] Cohort: We used data from patients with follow-up information, including survival status and time to event (death or censoring). Survival time was calculated based on the time from the date of diagnosis to death or last follow-up. Model: A risk score or predicted probability of survival was generated for each patient using the constructed Normo model. Time points: The survival prediction model was evaluated at three clinically relevant time points (3, 5, and 10 years after diagnosis).

[0066] (2) ROC curve construction:

[0067] To evaluate the ability of the model to distinguish between surviving and nonsurviving patients at each time point, a time-dependent ROC curve analysis was performed. At each time point, the survival probability (or risk score) was extracted from the model. The actual survival status (survival or death) of each patient at 3, 5, and 10 years was compared with the predicted probability generated by the model.

[0068] (3) Calculation of AUC (area under the curve):

[0069] AUC (Area Under the Curve): The area under the ROC curve (AUC) was calculated for each time point (3 years, 5 years, and 10 years). AUC is a summary indicator of the model's discriminatory ability. AUC interpretation: An AUC value of 0.5 indicates that the model has no discriminatory ability (equivalent to random guessing), while an AUC value of 1.0 indicates perfect discriminatory ability. AUC values ​​between 0.7 and 0.8 indicate acceptable discriminatory ability, between 0.8 and 0.9 indicate excellent discriminatory ability, and above 0.9 indicate outstanding discriminatory ability. 95% Confidence Interval (CI): AUC is reported as a 95% confidence interval to reflect the uncertainty of the estimate. Confidence intervals can be calculated using bootstrap (repeated random sampling) or other statistical methods.

[0070] (4) ROC curve drawing:

[0071] The ROC curves for 3-year, 5-year, and 10-year survival predictions were plotted, and the corresponding AUC values ​​were shown in the figure. The ROC curve shows the relationship between sensitivity (true positive rate) and 1-specificity (false positive rate) at different thresholds for predicting survival probability. The three ROC curves were color-coded according to the time point (blue, purple, etc. in the figure). Software and packages: ROC curve analysis used statistical software such as R or Python. In this model, the R language algorithm was used for analysis, and the timeROC or survivalROC packages were used to generate time-dependent ROC curves for survival data.

[0072] In addition, this test case also performed internal validation of the prognosis prediction model. The results are as follows: Figure 3 As shown, the C-index of the internal validation of the model is greater than 0.8 (left figure), and the normality of the C-index distribution is demonstrated by the QQ plot (right figure). Compared with the theoretical normal distribution, it shows that the C-index distribution conforms to the normal distribution assumption.

[0073] The specific steps are as follows:

[0074] (1) Data preparation:

[0075] The original dataset was used for model development and internal validation was performed using the Bootstrap method. Bootstrap sampling is a nonparametric bootstrap method that generates multiple repeated datasets by randomly sampling (with replacement) from the original dataset to evaluate the stability and predictive performance of the model. In each Bootstrap sampling, a sample of the same size as the original dataset is randomly drawn from the original dataset, but the sample can be repeated.

[0076] (2) Bootstrap sampling process:

[0077] Number of sampling: In this study, 500 Bootstrap samplings were performed. Each sampling generates a sample set of the same size as the original dataset (n = original sample size), in which some samples may be drawn multiple times and others may not be drawn. Model construction: For each Bootstrap sample set, the survival prognostic model was reconstructed and the predicted probability or risk score for each patient was calculated.

[0078] (3) C-index calculation: For each Bootstrap sample set, the C-index is calculated. The C-index value is used to measure the prediction accuracy of the model. The higher the C-index, the stronger the model's ability to distinguish.

[0079] (4) Display of C-index distribution: Distribution of C-index: The C-index values ​​of all 500 Bootstrap samples are summarized and displayed as a histogram, reflecting the distribution of the model's predictive performance on different sample sets. Confidence interval: In order to evaluate the stability of model performance, the median or mean of the C-index distribution is used as a summary indicator of model performance, and the 95% confidence interval (CI) of the C-index is also estimated by the distribution of Bootstrap samples. Typically, the confidence interval is constructed by sorting all Bootstrap C-index values ​​and taking the lower 5% and upper 95% values.

[0080] (5) Normality test: The normality of the C-exponential distribution is demonstrated by the QQ plot (right figure) and compared with the theoretical normal distribution. The QQ plot is used to check whether the C-exponential distribution meets the normal distribution assumption.

[0081] (6) Statistical software and parameters: This analysis used statistical software R or Python. In R, the boot package can be used to perform bootstrap sampling and calculate the C-index value; in Python, scikit-learn combined with the bootstrap method can achieve similar analysis.

[0082] In summary, the present invention has developed a dynamic prognostic nomogram for patients with essential thrombocythemia in the high-risk group of thrombosis, aiming to provide a more accurate risk assessment tool for clinical decision-making. The dynamic nomogram comprehensively considers multiple key prognostic factors, including age (divided into two groups: <60 years old and ≥60 years old), the presence or absence of splenomegaly, elevated white blood cell count, decreased hemoglobin level, and chromosome karyotype (abnormal or normal), in order to achieve a multi-dimensional assessment of the prognosis of patients with essential thrombocythemia in the high-risk group of thrombosis.

[0083] The applicant declares that the above is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Those skilled in the art should understand that any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention shall fall within the protection scope and disclosure scope of the present invention.

Claims

1. A method for establishing a prognosis prediction model for essential thrombocythemia patients in a high-risk group for thrombosis, characterized in that: The establishment method comprises the following steps: (1) Included medical records of patients with essential thrombocythemia who had a history of thrombosis; (2) Follow up the patients for a long time and collect clinical data during the follow-up period, including survival time, recurrence or death; (3) Use multiple imputation to handle missing values ​​in the data; (4) Screen out potential candidate variables related to disease prognosis from existing data based on literature and experience, and standardize or normalize numerical variables; (5) Perform univariate Cox regression analysis on each candidate variable to preliminarily evaluate the relationship between each variable and survival time or prognosis, obtain the regression coefficient and significance level. The regression coefficient represents the effect of the variable on the risk of event occurrence. Variables with a significance level P value less than 0.05 are selected as candidate variables for further multivariate regression analysis. (6) The significant variables screened out in the univariate Cox regression analysis were included in the multivariate Cox regression model to evaluate the independent prognostic effect of these variables after controlling other factors, and the regression coefficient and significance level were obtained. The regression coefficient represents the influence of each variable, and the variables with a significance level P value less than 0.05 were selected; (7) Based on the Cox regression results, the DynNom algorithm was used to generate a dynamic nomogram.

2. The establishment method according to claim 1, characterized in that: The multiple interpolation method in step (3) is a multiple interpolation method of chain equations; Preferably, the missing ratio imputed by the multiple imputation method in step (3) is ≤20%.

3. The establishment method according to claim 1 or 2, characterized in that: The candidate variables in step (4) include any one or a combination of at least two of age, gender, blood cell count, white blood cell count, hemoglobin count, biochemical parameters, spleen size by abdominal color Doppler ultrasound, bone marrow hematopoietic cell chromosome karyotype or JAK2 mutation.

4. The establishment method according to any one of claims 1 to 3, characterized in that: The variables with a significance level P value less than 0.05 in step (6) include whether the age is less than 60 years old, whether the spleen is enlarged, whether the chromosome karyotype is normal, any one of the white blood cell count and hemoglobin count, or a combination of at least two of them.

5. The establishment method according to any one of claims 1 to 4, characterized in that: Step (6) also includes adjusting the model using AIC / BIC.

6. A prognosis prediction model for essential thrombocythemia patients in the high-risk group of thrombosis, characterized in that: The prognosis prediction model is constructed by the establishment method described in any one of claims 1 to 5.

7. The prognosis prediction model according to claim 6, characterized in that: The prognostic prediction model includes: Data collection module: used to collect case data of patients with essential thrombocythemia who have a history of thrombosis, conduct long-term follow-up of the patients, and collect clinical data during the follow-up period, including survival time, recurrence or death; Data processing module: used to handle missing values ​​in data using multiple imputation methods; Variable screening and processing module: used to screen out potential candidate variables related to disease prognosis from existing data based on literature and experience, and to standardize or normalize numerical variables; Univariate Cox regression analysis module: used to perform univariate Cox regression analysis on each candidate variable, preliminarily evaluate the relationship between each variable and survival time or prognosis, obtain regression coefficients and significance levels. The regression coefficients represent the impact of the variables on the risk of events. Variables with a significance level P value less than 0.05 are selected as candidate variables for further multivariate regression analysis. Multivariate Cox regression analysis module: used to incorporate the significant variables screened out in the univariate Cox regression analysis into the multivariate Cox regression model, evaluate the independent prognostic effects of these variables after controlling other factors, obtain the regression coefficient and significance level. The regression coefficient represents the influence of each variable, and select variables with a significance level P value less than 0.05; Model output module: used to generate dynamic nomograms based on Cox regression results using the DynNom algorithm.

8. The prognosis prediction model according to claim 6 or 7, characterized in that: The multiple interpolation method is a multiple interpolation method of chain equations; Preferably, the multiple imputation method has a missing proportion of ≤20%; Preferably, the candidate variables in the variable screening and processing module include any one or a combination of at least two of age, gender, blood cell count, white blood cell count, hemoglobin count, biochemical parameters, spleen size by abdominal color Doppler ultrasound, chromosome karyotype of bone marrow hematopoietic cells, or JAK2 mutation; Preferably, the variables with a significance level P value less than 0.05 in the multi-factor Cox regression analysis module include whether the age is less than 60 years old, whether the spleen is enlarged, whether the chromosome karyotype is normal, any one or a combination of at least two of the white blood cell count and hemoglobin count; Preferably, the multi-factor Cox regression analysis module also includes using an AIC / BIC adjustment model.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program or instructions, and when the computer program or instructions are executed by a processor, the steps in the establishment method according to any one of claims 1 to 5 are implemented.

10. A computer device, characterized in that: The computer device comprises a memory and a processor, wherein the memory stores a computer program or instructions, and when the computer program or instructions are executed by the processor, the steps in the establishment method according to any one of claims 1 to 5 are implemented.

Images