CN114864097A - Method and device for establishing aorta dissection patient postoperative death prediction model - Google Patents
Method and device for establishing aorta dissection patient postoperative death prediction model Download PDFInfo
- Publication number
- CN114864097A CN114864097A CN202210250016.2A CN202210250016A CN114864097A CN 114864097 A CN114864097 A CN 114864097A CN 202210250016 A CN202210250016 A CN 202210250016A CN 114864097 A CN114864097 A CN 114864097A
- Authority
- CN
- China
- Prior art keywords
- risk factors
- group
- prediction model
- regression analysis
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Data Mining & Analysis (AREA)
- Biomedical Technology (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention discloses a method and a device for establishing a postoperative death prediction model of an aortic dissection patient, the prediction model is used for predicting the death probability of the acute aortic dissection patient within 30 days after the surgical operation, the method adopts the steps of collecting information data, grouping, carrying out regression analysis in groups, carrying out cross regression analysis among groups and carrying out multi-factor regression analysis to obtain independent risk factors, and assigning values based on the regression coefficients of the independent risk factors to establish a nomogram prediction model, the independent risk factors obtained by screening in the steps not only effectively simplify the independent risk factors, but also can fully screen out the potential independent risk factors, effectively improve the accuracy of the nomogram prediction model, the application range of the method is wider, so that the model establishing method disclosed by the invention has important significance in statistical analysis and screening, and related devices of the prediction model can be greatly convenient to establish and use the prediction model.
Description
Technical Field
The invention belongs to the field of biomedicine, and particularly relates to a method and a device for establishing a postoperative death prediction model for an aortic dissection patient.
Background
Aortic dissection surgical risk assessment is a key link for identifying high-risk patients, reducing the operative mortality rate and improving the surgical curative effect. Especially in acute StanfordA-type sandwich patients, the surgical risk is significantly increased and an accurate pre-operative risk factor assessment model is more required. At present, few models of death risk scoring after aortic dissection surgery at home and abroad exist, and the GeRAADA scoring is available internationally, but the scoring is mainly used for predicting death after pure ascending aorta replacement surgery and semi-arch replacement surgery, and the common surgery at home is full aortic arch replacement combined stent elephant nose surgery. Researches find that the GERADA score cannot accurately predict the death of patients with aortic dissection type A after full aortic arch combined stent-trunk operation, and the main defects of the existing prediction model (GERADA score) are as follows: (1) the operation mode accepted by the population establishing the model is different from the operation mode commonly used in China, and the postoperative mortality of the full aortic arch replacement combined bracket trunk operation cannot be accurately predicted. (2) The model cannot accurately predict the death rate after aortic dissection operation. Therefore, how to establish a prediction model which has wider application range and more accurate prediction and is used for the surgical death risk of the aortic dissection patient has great clinical significance.
Disclosure of Invention
The invention aims to provide a method and a device for establishing a prediction model of postoperative death of an aortic dissection patient, the prediction model is used for predicting the death of the acute aortic dissection patient within 30 days after a surgical operation, can more accurately evaluate the death risk of the acute aortic dissection patient within 30 days after the surgical operation, and is a prediction model with a wider application range.
In order to realize the purpose, the invention adopts the following technical scheme:
a method for establishing a prediction model of postoperative death of an aortic dissection patient is disclosed, wherein the prediction model is a nomogram prediction model which is used for predicting the death probability of the acute aortic dissection patient within 30 days after a surgical operation, and the method comprises the steps of obtaining independent risk factors through information data collection, grouping, intra-group regression analysis, inter-group cross regression analysis and multi-factor regression analysis, assigning values based on regression coefficients of the independent risk factors, and establishing the nomogram prediction model.
Further, the establishing method comprises the following steps:
s1, information data collection: collecting risk factors of acute aortic dissection concurrent surgical cases and survival information data within 30 days after surgery;
s2, grouping: randomly dividing the cases into a building module group and a verification group, and dividing the risk factors into preoperative group risk factors and intraoperative group risk factors;
s3, intraclass regression analysis: performing single-factor logistic regression analysis on the preoperative group risk factors by using the information data of the modeling group of S2 to obtain initial preoperative related risk factors, performing the rest preoperative group indirect related risk factors, performing the single-factor logistic regression analysis on the intraoperative group risk factors to obtain initial intraoperative related risk factors, and performing the rest intraoperative group indirect related risk factors;
s4, interclass cross regression analysis: respectively taking each initial preoperative relevant risk factor of S3 as a dependent variable, performing single-factor logistic regression analysis on the intraoperative indirect relevant risk factors of S3 to obtain newly-increased intraoperative relevant risk factors, respectively taking each initial intraoperative relevant risk factor of S3 as a dependent variable, and performing single-factor logistic regression analysis on the preoperative indirect relevant risk factors of S3 to obtain newly-increased preoperative relevant risk factors;
s5, multi-factor regression analysis: merging the initial preoperative relevant risk factors of S3, the initial intraoperative relevant risk factors of S3, the newly-increased preoperative relevant risk factors of S4 and the newly-increased intraoperative relevant risk factors of S4, bringing the merged factors into a multi-factor logistic regression model, and screening out independent risk factors by adopting a stepwise regression method;
s6, model establishment and verification: and assigning values to the independent risk factors based on the multi-factor logistic regression model of S5, establishing a nomogram prediction model for predicting death within 30 days after the operation, and performing model verification.
Further, all categorical variables involved in the method are expressed as percentages and the continuous variables are expressed as standard deviations.
Further, the continuous variable adopts a t test or a wilcoxon rank sum test, and the classified variable adopts a chi-square test.
Further, the initial preoperative related risk factors of S3 include: hemodynamically unstable, persistent interbedded-related abdominal pain, estimated glomerular filtration rate of less than 50ml/min, S3 said initial intraoperative related risk factors including: combining CABG surgery and semi-arch replacement surgery.
Further, the newly added preoperative related risk factors of S4 include: chronic kidney disease, CT indicates that the abdominal cavity is not well perfused, coronary artery disease, left ventricle has the diastolic diameter less than 45mm, a large amount of pericardial effusion and descending aorta true cavity severe stenosis, S4 the related risk factors in the newly-increased operation include: extracorporeal circulation time greater than 4 hours, aortic valve replacement, aortic arch replacement.
Further, the independent risk factors of S5 include: the pre-operative left ventricular end diastolic diameter is less than 45mm, the estimated glomerular filtration rate is less than 50ml/min, abdominal pain associated with the persisting interlayer, dry perfusion failure of the abdominal cavity suggested by CT, combined CABG surgery and extracorporeal circulation time is greater than 4 hours.
An aortic dissection patient postoperative death prediction model creation apparatus, the apparatus comprising an input and an output, the apparatus further comprising:
the patient information acquisition module is used for screening and acquiring patient data meeting preset grouping conditions and grouping the patient risk factors according to the preset conditions;
the first analysis module is used for performing single-factor regression analysis on each group of grouped risk factors respectively, and each group of risk factors are analyzed to obtain initial related risk factors and indirect related risk factors respectively;
the analysis module II is used for carrying out correlation regression analysis on each initial relevant risk factor in each group and all indirect relevant risk factors in the other group respectively based on each group of initial relevant risk factors and indirect relevant risk factors obtained by the analysis module I, and screening new risk factors from the indirect relevant risk factors in each group;
the analysis module III is used for combining all initial relevant risk factors acquired by the analysis module I and newly increased risk factors acquired by the analysis module II, bringing the combined initial relevant risk factors and newly increased risk factors into a multi-factor logistic regression model, and screening out independent risk factors by adopting a stepwise regression method;
and the model establishing module is used for assigning the independent risk factors and establishing a prediction model based on the multi-factor logistic regression model of the analysis module III.
An aortic dissection patient postoperative death prediction model building terminal device, the terminal device comprises an interface end, the terminal device further comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, and the processor executes the computer program to realize the aortic dissection patient postoperative death nomogram prediction model building method.
A computer-readable storage medium, the storage medium including an interface plug, the computer-readable storage medium further including a stored computer program, wherein when the computer program is executed, the apparatus in which the computer-readable storage medium is located is controlled to execute the method for building a postoperative death nomogram prediction model for an aortic dissection patient.
The invention discloses a method and a device for establishing a post-operation death prediction model of an aortic dissection patient, which have the following beneficial effects:
(1) the method for establishing the nomogram prediction model gradually screens and confirms 6 independent risk factors through analysis methods such as information data collection, grouping, intra-group regression analysis, inter-group cross regression analysis, multi-factor regression analysis and the like, and comprises the following steps of: through the intraclass regression analysis, 3 kinds of initial preoperative relevant risk factors and 2 kinds of initial intraoperative relevant risk factors are screened and confirmed, further through interclass cross regression analysis, 3 kinds of new intraoperative relevant risk factors and 6 kinds of new preoperative relevant risk factors are obtained, and six kinds of independent risk factors are obtained through the subsequent multifactor regression analysis screening simplification, in the process, three kinds of final 6 independent risk factors are obtained through intraclass regression analysis screening: estimating glomerular filtration rate less than 50ml/min, persistent interbed-related abdominal pain and combined CABG surgery, and further screening by interclass cross regression analysis to obtain the other three of the final 6 independent risk factors by using the results of the intraclass regression analysis: the left ventricular end diastolic diameter before operation is less than 45mm, CT indicates that the abdominal cavity is not well perfused and the extracorporeal circulation time is more than 4 hours, and the other 8 risk factors are obtained simultaneously when the independent risk factors are obtained through the intragroup regression analysis and the interclass cross regression analysis: the method for establishing the nomogram prediction model has the advantages that hemodynamics are unstable, the hemiarcade replacement surgery, chronic kidney diseases, coronary artery diseases, a large amount of pericardial effusion, descending aorta true lumen severe stenosis, aortic valve replacement and aortic arch replacement are carried out, and 8 risk factors are further screened through subsequent multi-factor regression analysis to obtain the final 6 independent risk factors.
(2) The nomogram prediction model established by the method is a simple and effective tool, the death incidence rate of 30 days after the acute aortic dissection full aortic arch replacement surgery can be accurately predicted, only 6 independent risk factors convenient to acquire clinically are included in the prediction model, the number of the predicted risk factors is obviously less than that of a geraada model, a Euroscore II model, an STS model and the like, so that the prediction factors are easier to acquire, the applicable population range is wider, and compared with other current prediction models, the nomogram prediction model is higher in accuracy as shown by verification results.
(3) The method for establishing the nomogram prediction model is used for innovatively screening independent risk factors of which the left ventricular end diastolic diameter is smaller than 45mm and is also used for operative death, the independent risk factors are obtained through grouping, intra-group regression analysis and inter-group cross regression analysis step by step and finally through inter-group cross regression analysis screening, and the method for establishing the nomogram prediction model is further proved to have remarkable beneficial effects in the aspect of screening and searching related factors and have important significance in statistical analysis and screening.
(4) The invention discloses a plurality of devices for establishing a prediction model, which comprise: the device greatly simplifies the process of establishing a relevant prediction model by using the prediction model establishing method disclosed by the invention on one hand, and simultaneously enables the use of the prediction model to be more convenient, and can be used for establishing attempts of other disease prediction models by adjusting preset conditions in the device, so that the device has wider application range, and the device has important inspiration significance for establishing other disease prediction models.
Drawings
FIG. 1 is a schematic illustration of a nomogram prediction model in accordance with the present invention;
FIG. 2 is a first calibration curve of the alignment chart of the present invention;
FIG. 3 is a first schematic diagram of the ROC curve of the present invention;
FIG. 4 is a second calibration curve of the nomogram of the present invention;
FIG. 5 is a second schematic view of the ROC curve of the present invention;
FIG. 6 is a schematic diagram showing comparison of brier scores according to the present invention;
FIG. 7 is a graph comparing AUC values (c statistic) of the present invention;
FIG. 8 is a schematic flow chart of a method for establishing a nomogram predictive model in accordance with the present invention;
FIG. 9 is a schematic block diagram of an apparatus for modeling a post-operative death prediction model for an aortic dissection patient;
FIG. 10 is a schematic diagram of a terminal device module for establishing a model for predicting postoperative death of an aortic dissection patient;
fig. 11 is a schematic diagram of a computer-readable storage medium module.
Detailed Description
The invention discloses a method and a device for establishing an aortic dissection patient postoperative death prediction model, and relates to an aortic dissection patient postoperative death prediction model, wherein the nomogram prediction model relates to independent risk factors and comprises the following steps: the left ventricular end diastolic diameter before operation is less than 45mm, the estimated glomerular filtration rate is less than 50ml/min, abdominal pain related to interlayer continuously exists, dry abdominal perfusion failure is prompted by CT, CABG operation and extracorporeal circulation time are combined for more than 4 hours, the independent risk factors are brought into the nomogram prediction model, and the nomogram prediction model is used for death prediction of patients suffering from acute aortic interlayer within 30 days after surgical operation.
A method for establishing a prediction model of postoperative death of an aortic dissection patient is used for predicting the death probability of the acute aortic dissection patient within 30 days after a surgical operation, and the method obtains independent risk factors through information data collection, grouping, intra-group regression analysis, inter-group cross regression analysis and multi-factor regression analysis, assigns values based on regression coefficients of the independent risk factors, and establishes the prediction model of the nomogram.
Further, the establishing method comprises the following steps: (see FIG. 8)
S1, information data collection: collecting risk factors of acute aortic dissection concurrent surgical cases and survival information data within 30 days after surgery;
s2, grouping: randomly dividing the cases into a building module group and a verification group, and dividing the risk factors into preoperative group risk factors and intraoperative group risk factors;
s3, intraclass regression analysis: performing single-factor logistic regression analysis on the preoperative group risk factors by using the information data of the modeling group of S2 to obtain initial preoperative related risk factors, performing the rest preoperative group indirect related risk factors, performing the single-factor logistic regression analysis on the intraoperative group risk factors to obtain initial intraoperative related risk factors, and performing the rest intraoperative group indirect related risk factors;
s4, interclass cross regression analysis: respectively taking each initial preoperative relevant risk factor of S3 as a dependent variable, performing single-factor logistic regression analysis on the intraoperative indirect relevant risk factors of S3 to obtain newly-increased intraoperative relevant risk factors, respectively taking each initial intraoperative relevant risk factor of S3 as a dependent variable, and performing single-factor logistic regression analysis on the preoperative indirect relevant risk factors of S3 to obtain newly-increased preoperative relevant risk factors;
s5, multi-factor regression analysis: merging the initial preoperative relevant risk factors of S3, the initial intraoperative relevant risk factors of S3, the newly-increased preoperative relevant risk factors of S4 and the newly-increased intraoperative relevant risk factors of S4, bringing the merged factors into a multi-factor logistic regression model, and screening out independent risk factors by adopting a stepwise regression method;
s6, model establishment and verification: and assigning values to the independent risk factors based on the multi-factor logistic regression model of S5, establishing a nomogram prediction model for predicting death within 30 days after the operation, and performing model verification.
Further, all categorical variables involved in the method are expressed as percentages and the continuous variables are expressed as standard deviations.
Further, the continuous variable adopts a t test or a wilcoxon rank sum test, and the classified variable adopts a chi-square test.
Further, in S2, the ratio of the number of the building groups to the number of the verification groups is 7: 3.
further, the initial preoperative related risk factors of S3 include: hemodynamically unstable, persistent interbedded-related abdominal pain, estimated glomerular filtration rate of less than 50ml/min, S3 said initial intraoperative related risk factors including: combining CABG surgery and semi-arch replacement surgery.
Further, S4 shows the new preoperative related risk factors including: chronic kidney disease, CT indicates that the abdominal cavity is not well perfused, coronary artery disease, left ventricle has the diastolic diameter less than 45mm, a large amount of pericardial effusion and descending aorta true cavity severe stenosis, S4 the related risk factors in the newly-increased operation include: extracorporeal circulation time greater than 4 hours, aortic valve replacement, aortic arch replacement.
Further, the independent risk factors of S5 include: the pre-operative left ventricular end diastolic diameter is less than 45mm, the estimated glomerular filtration rate is less than 50ml/min, abdominal pain associated with the persisting interlayer, dry perfusion failure of the abdominal cavity suggested by CT, combined CABG surgery and extracorporeal circulation time is greater than 4 hours.
Further, a t test is adopted when the continuous variable accords with positive distribution, and a wilcoxon rank sum test is adopted when the continuous variable does not accord with positive distribution.
Further, all statistical analyses were performed using rstudio4.0.2 software.
An aortic dissection patient postoperative death prediction model creation apparatus, the apparatus comprising an input and an output, the apparatus further comprising:
the patient information acquisition module is used for screening and acquiring patient data meeting preset grouping conditions and grouping the patient risk factors according to the preset conditions;
the first analysis module is used for performing single-factor regression analysis on each group of grouped risk factors respectively, and each group of risk factors are analyzed to obtain initial related risk factors and indirect related risk factors respectively;
the analysis module II is used for carrying out correlation regression analysis on each initial relevant risk factor in each group and all indirect relevant risk factors in the other group respectively based on each group of initial relevant risk factors and indirect relevant risk factors obtained by the analysis module I, and screening new risk factors from the indirect relevant risk factors in each group;
the analysis module III is used for combining all initial relevant risk factors acquired by the analysis module I and newly increased risk factors acquired by the analysis module II, bringing the combined initial relevant risk factors and newly increased risk factors into a multi-factor logistic regression model, and screening out independent risk factors by adopting a stepwise regression method;
and the model establishing module is used for assigning the independent risk factors and establishing a prediction model based on the multi-factor logistic regression model of the analysis module III.
An aortic dissection patient postoperative death prediction model building terminal device, the terminal device comprises an interface end, the terminal device further comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, and the processor executes the computer program to realize the aortic dissection patient postoperative death nomogram prediction model building method.
A computer-readable storage medium, the storage medium including an interface plug, the computer-readable storage medium further including a stored computer program, wherein when the computer program is executed, the apparatus in which the computer-readable storage medium is located is controlled to execute the method for building a postoperative death nomogram prediction model for an aortic dissection patient.
Example 1
1156 patients of acute aortic dissection concurrent surgical operation of the hospital extrafuds in 2010-1 month to 2020-12 months are continuously collected and screened, all the patients are subjected to full-aortic CTA examination (scanning range from the head-arm blood vessel to the bilateral femoral artery level) to be clearly diagnosed before operation, the abnormality of the heart structure and function is clearly determined Through Thoracic Echocardiography (TTE), and coronary artery CTA is clearly determined as coronary lesion. Demographic data, preoperative risk factors, important intraoperative information, and 30-day postoperative survival were collected for all patients.
Demographic data, preoperative risk factors, important intraoperative information for all patients included: age, gender, body mass index, Marfan's syndrome, diabetes, stroke, chronic kidney disease, chronic obstructive pulmonary disease, elevated glutamate pyruvate transaminase, poor kidney perfusion, grade 3 or 4 of New York Heart function, hypertension, coronary artery disease, atrial fibrillation, previous cardiovascular surgery, previous aortic intervention, hemodynamically unstable, left ventricular end diastolic diameter less than 45mm, decreased ejection fraction, severe aortic valve regurgitation, massive pericardial effusion, severe luminal narrowing of descending aorta, persistent interbedded related abdominal pain, lower extremity arterial ischemic symptoms, coronary ostial involvement, carotid ostial involvement, CT-suggested celiac trunk perfusion, superior mesenteric artery perfusion, estimated glomerular filtration rate less than 50ml/min, imaging-suggested iliac femoral perfusion, combined aortic root surgery, coronary artery disease, peripheral arterial thrombosis, vascular occlusion of arterial thrombosis, vascular occlusion of cardiac vascular occlusion, vascular occlusion of a, Carotid artery bypass graft surgery, combined CABG surgery, extracorporeal circulation time longer than 4 hours, deep low temperature circulation stopping temperature, deep low temperature circulation stopping time, aortic valve replacement, semi-arch replacement, aortic arch replacement, axillary artery bypass graft surgery, ascending aorta-femoral artery bypass graft surgery and ascending aorta-abdominal aorta bypass graft surgery, wherein the total risk factors are 42.
Example 2
Risk factors were grouped, and all cases were as follows 7: 3, randomly dividing the risk factors into a building module (n is 806) and a verification module (n is 350), dividing the risk factors into preoperative group risk factors and intraoperative group risk factors in 42, wherein the preoperative group risk factors comprise: age, gender, body mass index, Marfan's syndrome, diabetes, stroke, chronic kidney disease, chronic obstructive pulmonary disease, elevated glutamate pyruvate transaminase, poor kidney perfusion, grade 3 or 4 of New York Heart function, hypertension, coronary artery disease, atrial fibrillation, previous cardiovascular surgery, previous aortic intervention, hemodynamically unstable, left ventricular end diastolic diameter less than 45mm, decreased ejection fraction, severe aortic valve regurgitation, massive pericardial effusion, severe luminal narrowing of descending aorta, persistent dissection-related abdominal pain, lower extremity arterial ischemic symptoms, coronary ostial involvement, carotid ostial involvement, CT-suggested celiac trunk perfusion, superior mesenteric artery perfusion, estimated glomerular filtration rate less than 50ml/min, imaging-suggested iliac femoral perfusion, combined aortic root surgery, there are a total of 31 preoperative groups of risk factors, including: carotid artery bypass graft surgery, combined CABG surgery, extracorporeal circulation time longer than 4 hours, deep low temperature circulation stopping temperature, deep low temperature circulation stopping time, aortic valve replacement, semi-arch replacement, aortic arch replacement, axillary artery bypass graft surgery, ascending aorta-femoral artery bypass graft surgery and ascending aorta-abdominal aorta bypass graft surgery, wherein 11 intraoperative risk factors are combined.
Example 3
All classification variables involved in the process of establishing the prediction model are expressed as percentage percent, and continuous variables are expressed as standard deviation. And (3) screening all preoperative and intraoperative risk factors possibly dying within 30 days after operation by using single factor analysis in the building module, adopting t test when continuous variables are compositely in accordance with positive distribution, adopting wilcoxon rank sum test when the continuous variables are not in accordance with the positive distribution, adopting chi-square test on classified variables, and obtaining the risk factors with single factor analysis positive (P <0.1) through regression analysis.
Performing intraclass regression analysis, using information of modeling module, classifying risk factors (age, sex, body mass index, Marfan syndrome, diabetes, stroke, chronic kidney disease, chronic obstructive pulmonary disease, increased glutamate pyruvate transaminase, renal hypoperfusion, grade 3 or 4 of New York Heart function, hypertension, coronary artery disease, atrial fibrillation, cardiovascular surgery, aorta intervention, hemodynamics instability, left ventricular diastolic diameter less than 45mm, ejection fraction reduction, severe aortic valve regurgitation, massive pericardial effusion, severe stenosis of descending aorta true lumen, abdominal pain related to persistent interlayer, ischemic symptom of lower extremity artery, coronary artery opening affected, carotid artery opening affected, CT (computed tomography) prompting of hypoperfusion, supramesenteric artery hypoperfusion, estimated glomerular filtration rate less than 50ml/min, blood pressure, blood flow, blood pressure, blood flow, blood pressure, blood flow, blood pressure, blood flow, blood pressure, blood flow, blood pressure, blood flow, blood pressure, blood, Imaging prompts that iliac-femoral artery is poorly perfused and aorta root surgery is combined, 31 preoperative group risk factors are counted), and single-factor logistic regression analysis is carried out on the death correlation within 30 days after surgery, 3 initial preoperative related risk factors are obtained through analysis, and the 3 initial preoperative related risk factors comprise: hemodynamically unstable, persistent dissecting-related abdominal pain, estimated glomerular filtration rate of less than 50ml/min, and the remaining 28 non-directly related risk factors for the preoperative group, and performing one-way logistic regression analysis on the intraoperative group risk factors (carotid bypass graft, combined CABG surgery, extracorporeal circulation time of more than 4 hours, deep hypothermic resting circulation temperature, deep hypothermic resting circulation time, aortic valve replacement, hemi-arch replacement, aortic arch replacement, axillary artery bypass graft, ascending aorta-femoral bypass graft, ascending aorta-abdominal aorta bypass graft, total 11 intraoperative group risk factors) and postoperative 30-day mortality correlation, and analyzing to obtain 2 initial intraoperative related risk factors, wherein the 2 initial intraoperative related risk factors comprise: combining CABG surgery and hemi-arch replacement, the remaining 9 were intraoperative group non-directly related risk factors.
Example 4
Performing interclass cross regression analysis on the 9 intraoperative groups of indirectly related risk factors (carotid artery bypass graft, extracorporeal circulation time longer than 4 hours, deep hypothermia shutdown temperature, deep hypothermia shutdown circulation time, aortic valve replacement, aortic arch replacement, axillary artery bypass graft, ascending aorta-femoral artery bypass graft, ascending aorta-abdominal aorta bypass graft) by using the 3 initial preoperative related risk factors (hemodynamically unstable, persistently existing interlayer related abdominal pain, estimated glomerular filtration rate less than 50ml/min) as dependent variables respectively to obtain 3 new intraoperative related risk factors, wherein the 3 new intraoperative related risk factors comprise: extracorporeal circulation time greater than 4 hours, aortic valve replacement, aortic arch replacement, wherein the 2 initial intraoperative associated risk factors (combined CABG surgery, semi-arch replacement) are used as dependent variables, and 28 preoperative non-directly related risk factors (age, sex, body mass index, Marfan syndrome, diabetes, past stroke, chronic kidney disease, chronic obstructive pulmonary disease, elevated glutamate pyruvate transaminase, renal hypoperfusion, grade 3 or 4 New York Heart function, hypertension, coronary artery disease, atrial fibrillation, past cardiovascular surgery, past aortic intervention, left ventricular diastolic diameter less than 45mm, ejection fraction reduction, severe aortic valve regurgitation, large pericardial effusion, descending aorta true lumen severe lower extremity, arterial ischemic symptoms, coronary artery opening involvement, carotid artery opening involvement, and carotid artery opening involvement) are used as dependent variables, CT-prompt abdominal trunk perfusion failure, superior mesenteric artery perfusion failure, imaging-prompt iliofemoral artery perfusion failure, combined aortic root surgery) to perform single-factor logistic regression analysis, to obtain 6 newly-added preoperative relevant risk factors, including 6 newly-added preoperative relevant risk factors: chronic kidney disease and CT indicate that abdominal cavity is not well perfused, coronary artery disease, left ventricle diastolic diameter is less than 45mm, a large amount of pericardial effusion and descending aorta true cavity severe stenosis.
Example 5
The multi-factor regression analysis and prediction model establishment and verification are characterized in that 3 initial preoperative relevant risk factors (hemodynamics instability, persistent interlayer-related abdominal pain, estimated glomerular filtration rate less than 50ml/min), 2 initial intraoperative relevant risk factors (combination of CABG operation and hemi-bow replacement operation), 6 newly added preoperative relevant risk factors (chronic renal disease, CT (computed tomography) prompt abdominal dry perfusion failure, coronary artery disease, left ventricular diastolic end diameter less than 45mm, a large amount of pericardial effusion, descending aorta true lumen severe stenosis) and 3 newly added intraoperative relevant risk factors (extracorporeal circulation time longer than 4 hours, aortic valve replacement and aortic arch replacement) are combined and are brought into a multi-factor logistic regression model together, and a stepwise regression method is adopted to screen out independent risk factors, wherein the independent risk factors comprise: the pre-operative left ventricular end diastolic diameter is less than 45mm, the estimated glomerular filtration rate is less than 50ml/min, abdominal pain associated with the persisting interlayer, dry perfusion failure of the abdominal cavity suggested by CT, combined CABG surgery and extracorporeal circulation time is greater than 4 hours.
And then, assigning values based on the regression coefficients of the independent risk factors in the 6 groups by using a logistic regression model, and establishing a nomogram prediction model equation for predicting death within 30 days after operation: p (t) ═ λ 0 (t)exp(β 1 x 1 +β 2 x 2 +…+β k x k ) The nomogram prediction model is shown in figure 1, the corresponding weights and coefficients of 6 independent risk factors are shown in table 1, and P (t) is the death incidence rate of the aortic dissection patients after operation, lambda 0 (t) is a function for calculating AKI, x 1 、x 2 ...x k As covariate, beta i Is x i The regression coefficient of (2).
TABLE 1 Risk factors and weights
And (3) model verification, namely evaluating the calibration degree of the nomogram by using a calibration curve and a brier score in a verification group, evaluating the discrimination degree of the model by using an ROC curve and an AUC value (c statistic, 95% CI) thereof, and in a modeling data set, the calibration curve of a nomogram is shown in figure 2, the brier score is 0.0523, the ROC curve is shown in figure 3, and the AUC value is 0.7851. In the validation set of data, the alignment curve of the nomogram is shown in fig. 4, its brier score is 0.0613, its ROC curve is shown in fig. 5, and its AUC value is 0.7819. Meanwhile, 4 common machine learning algorithms (NB, SVM, randomforest and XGboost) are selected to be modeled by using modeling group data, then verification is carried out in verification group data, brier scores and AUC 95% CI of each ML algorithm model are calculated and compared with a nomogram model.
In the modeling group and the verification group, a calibreration curve (150 times of sampling) and brier scores are adopted to evaluate the calibration degree of the model, and the discrimination degree of the model is evaluated by using an ROC curve and an AUC value (c statistic and 95% confidence interval thereof). 4 common machine learning algorithms (naive Bayes, support vector machine, regression forest and extreme gradient lifting) are adopted in a modeling group to establish corresponding machine learning models, brier scores are used in a verification group to evaluate the calibration degree of the machine learning algorithm models, AUC values (c statistics and 95% confidence intervals) of the machine learning algorithm models are used to evaluate the discrimination degree of the models, and the discrimination degree is compared with a nomogram prediction model. All statistical analyses were performed using rstudio4.0.2 software.
Compared with other machine learning algorithm models, fig. 6 shows that brier scores of the nomogram model and other 4 machine learning algorithms in the verification group dataset are compared, and it can be seen that brier scores of the nomogram model are smaller than those of the other 4 machine learning algorithm models, which indicates that the degree of calibration is higher. Fig. 7 shows a comparison of the nomogram model with AUC values (c statistics) of other 4 machine learning algorithms in the verification group dataset, which shows that the AUC values of the nomogram model are larger than those of the other 4 machine learning algorithms, which indicates that the degree of discrimination is higher and the applicable population range is wider.
The invention innovatively finds that the left ventricular end diastolic diameter before the operation is less than 45mm is an independent risk factor for death after the operation, particularly persistent abdominal pain has strong prediction effect, and prompts that the adverse perfusion of abdominal organs before the operation has great influence on the prognosis of the total-arch replacement operation, the adverse perfusion of the abdominal organs is a destructive complication, can cause 70% -100% of death after the operation of an acute A-type interlayer, and prompts that people should evaluate the perfusion condition of the abdominal organs (symptom signs and imaging examination) of a patient as early as possible, and should perform reasonable and effective intervention as early as possible on an acute interlayer patient who does not have the adverse perfusion of the abdominal organs and has a tendency to progress.
The nomogram prediction model is a simple and effective prediction tool for predicting death within 30 days after acute aortic dissection full aortic arch replacement. The prediction model only incorporates 6 risk factors in preoperative operations which are convenient to acquire clinically, the number of the predicted risk factors is obviously less than that of a geraada model, a euroscoreII model, an STS model and the like, the prediction factors are easier to acquire and more accurate to predict, the applicable population range is wider, and in addition, the using convenience of the nomogram is higher than that of a webpage calculator, and the prediction model can be used as a reliable bedside using tool.
The nomogram prediction model is a simple and effective tool, and can accurately predict the 30-day death incidence rate after acute aortic dissection full aortic arch replacement surgery.
Example 6
An aortic dissection patient postoperative death prediction model building device, the device further comprises:
the patient information acquisition module is used for screening the acquired patient data to obtain patient data meeting preset grouping conditions and grouping the patient risk factors according to the preset conditions; screening a patient data information base according to whether the patient is an aortic dissection patient or not to obtain an aortic dissection patient information base, and classifying risk factors related in the aortic dissection patient information base into preoperative group risk factors and intraoperative group risk factors;
the first analysis module is used for performing single-factor regression analysis on each group of grouped risk factors respectively, and each group of risk factors are analyzed to obtain initial risk factors and non-initial risk factors respectively; performing single-factor logistic regression analysis on the acquired preoperative group risk factors to obtain initial preoperative related risk factors, performing the rest preoperative group indirect related risk factors, performing the single-factor logistic regression analysis on the intraoperative group risk factors to obtain initial intraoperative related risk factors, and performing the rest intraoperative group indirect related risk factors;
the analysis module II is used for performing correlation regression analysis on the initial risk factors in each group and all the non-initial risk factors in the other group on the basis of the initial risk factors and the non-initial risk factors in each group obtained by the analysis module I, and screening new risk factors from the non-initial risk factors in each group; respectively taking initial preoperative related risk factors acquired from the analysis module I as dependent variables, performing single-factor logistic regression analysis on the intraoperative group of indirect related risk factors acquired from the analysis module I to obtain new intraoperative related risk factors, respectively taking each initial intraoperative related risk factor acquired from the analysis module I as a dependent variable, and performing single-factor logistic regression analysis on the preoperative group of indirect related risk factors acquired from the analysis module I to obtain new preoperative related risk factors;
the analysis module III is used for combining all initial risk factors in the analysis module I and newly increased left and right risk factors in the analysis module II, bringing the combined initial risk factors and newly increased left and right risk factors into a multi-factor logistic regression model, and screening out independent risk factors by adopting a stepwise regression method; combining the initial preoperative relevant risk factors and the initial intraoperative relevant risk factors acquired by the analysis module I, and the newly added preoperative relevant risk factors and the newly added intraoperative relevant risk factors acquired by the analysis module II, bringing the combined factors into a multi-factor logistic regression model, and screening out independent risk factors by adopting a stepwise regression method
And the model establishing module is used for assigning the independent risk factors and establishing a prediction model based on the multi-factor logistic regression model of the analysis module III.
FIG. 9 is a schematic diagram of a model building apparatus for predicting postoperative death of an aortic dissection patient, in which a patient information acquisition module acquires patient information via an input terminal, acquires patient information by setting disease conditions and screening, and groups all risk factors; the first analysis module acquires grouping information from the patient information acquisition module, and performs logistic single-factor regression analysis (namely the regression analysis in the groups) on the relevance of each group of risk factors and the death after the disease operation to respectively acquire initial relevant risk factors and indirect relevant risk factors; the analysis module II acquires the analysis result information from the analysis module I and further performs interclass cross regression analysis to obtain newly added risk factors of each group; the analysis module III acquires initial relevant risk factors from the analysis module I, acquires newly increased risk factors from the analysis module II, further performs multi-factor logistic regression analysis, and screens out independent risk factors by adopting a stepwise regression method; and the model establishing module is used for assigning the independent risk factors based on the multi-factor logistic regression model of the analysis module III, establishing a prediction model and feeding back the prediction model to a user through an output end.
Example 7
An aortic dissection patient postoperative death prediction model building terminal device comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein the processor realizes the processes of data information acquisition, intra-group regression analysis, inter-group cross regression analysis, multi-factor regression analysis and prediction model building when executing the computer program.
Fig. 10 is a schematic diagram of a terminal device module for establishing a post-operation death prediction model for aortic dissection patients, the device mainly includes a memory and a processor, the memory is configured with a computer program, the processor acquires an analysis signal through an interface terminal, the computer program configured in the memory is operated, the computer program analyzes and processes patient information data transmitted through the interface terminal, and the analysis and processing process includes the processes of data information screening acquisition, intra-group regression analysis, inter-group cross regression analysis, multi-factor regression analysis, prediction model establishment and the like.
Example 8
A computer-readable storage medium comprising a stored computer program, wherein the computer program, when executed, controls a device in which the computer-readable storage medium is located to perform data information acquisition, intra-group regression analysis, inter-group cross regression analysis, multi-factor regression analysis, and predictive model building processes.
Fig. 11 is a schematic diagram of a computer-readable storage medium module, which includes an interface plug and a readable storage medium, where a computer program is installed in a memory scale storage medium, the readable storage medium is connected to an external device through the interface plug, the readable storage medium obtains patient data information from the external device through the interface plug, and starts the computer program installed in the readable storage medium, analyzes the obtained patient data information, including patient data information screening, intra-group regression analysis, inter-group cross regression analysis, multi-factor regression analysis, prediction model building, and the like, and feeds back the prediction model to a user through the external device.
The present invention is further illustrated and described in the above embodiments, which are only used to help understand the method and the core idea of the present application, and the content of the present specification should not be construed as limiting the present application since the skilled person can change the specific implementation and application scope according to the idea of the present application.
Claims (10)
1. A method for establishing a prediction model of postoperative death of an aortic dissection patient, wherein the prediction model is a nomogram prediction model which is used for predicting the death probability of the acute aortic dissection patient within 30 days after a surgical operation, and is characterized in that the method obtains independent risk factors through information data collection, grouping, intra-group regression analysis, inter-group cross regression analysis and multi-factor regression analysis, and assigns values based on regression coefficients of the independent risk factors to establish the nomogram prediction model.
2. The method for building a model for predicting postoperative death of an aortic dissection patient according to claim 1, wherein the building method comprises the following steps:
s1, information data collection: collecting risk factors of acute aortic dissection concurrent surgical cases and survival information data within 30 days after surgery;
s2, grouping: randomly dividing the cases into a building module group and a verification group, and dividing the risk factors into preoperative group risk factors and intraoperative group risk factors;
s3, intraclass regression analysis: performing single-factor logistic regression analysis on the preoperative group risk factors by using the information data of the modeling group of S2 to obtain initial preoperative related risk factors, performing the rest preoperative group indirect related risk factors, performing the single-factor logistic regression analysis on the intraoperative group risk factors to obtain initial intraoperative related risk factors, and performing the rest intraoperative group indirect related risk factors;
s4, interclass cross regression analysis: respectively taking each initial preoperative relevant risk factor of S3 as a dependent variable, performing single-factor logistic regression analysis on the intraoperative group indirect relevant risk factors of S3 to obtain new intraoperative relevant risk factors, respectively taking each initial intraoperative relevant risk factor of S3 as a dependent variable, and performing single-factor logistic regression analysis on the preoperative group indirect relevant risk factors of S3 to obtain new preoperative relevant risk factors;
s5, multi-factor regression analysis: merging the initial preoperative relevant risk factors of S3, the initial intraoperative relevant risk factors of S3, the newly-increased preoperative relevant risk factors of S4 and the newly-increased intraoperative relevant risk factors of S4, bringing the merged factors into a multi-factor logistic regression model, and screening out independent risk factors by adopting a stepwise regression method;
s6, model establishment and verification: and assigning values to the independent risk factors based on the multi-factor logistic regression model of S5, establishing a nomogram prediction model for predicting death within 30 days after the operation, and performing model verification.
3. The method of claim 2, wherein all the classification variables are expressed as percentage and the continuous variable is expressed as standard deviation.
4. The method for establishing the post-operative death prediction model for aortic dissection patients as claimed in claim 3, wherein the continuous variable is t-test or wilcoxon rank sum test, and the classification variable is chi-square test.
5. The method of claim 2, wherein the initial preoperative related risk factors of S3 include: hemodynamically unstable, persistent interbedded-related abdominal pain, estimated glomerular filtration rate of less than 50ml/min, S3 said initial intraoperative related risk factors including: combining CABG surgery and semi-arch replacement surgery.
6. The method of claim 2, wherein the step of creating the model for predicting postoperative death of aortic dissection patients comprises the step of S4: chronic kidney disease, CT indicates that the abdominal cavity is not well perfused, coronary artery disease, left ventricle has the diastolic diameter less than 45mm, a large amount of pericardial effusion and descending aorta true cavity severe stenosis, S4 the related risk factors in the newly-increased operation include: extracorporeal circulation time greater than 4 hours, aortic valve replacement, aortic arch replacement.
7. The method according to claim 2, wherein the independent risk factors of S5 include: the pre-operative left ventricular end diastolic diameter is less than 45mm, the estimated glomerular filtration rate is less than 50ml/min, abdominal pain associated with the persisting interlayer, dry perfusion failure of the abdominal cavity suggested by CT, combined CABG surgery and extracorporeal circulation time is greater than 4 hours.
8. An aortic dissection patient postoperative death prediction model creation device, the device comprising an input and an output, the device further comprising:
the patient information acquisition module is used for screening and acquiring patient data meeting preset grouping conditions and grouping the patient risk factors according to the preset conditions;
the first analysis module is used for performing single-factor regression analysis on each group of grouped risk factors respectively, and each group of risk factors are analyzed to obtain initial related risk factors and indirect related risk factors respectively;
the analysis module II is used for carrying out correlation regression analysis on each initial relevant risk factor in each group and all indirect relevant risk factors in the other group respectively based on each group of initial relevant risk factors and indirect relevant risk factors obtained by the analysis module I, and screening new risk factors from the indirect relevant risk factors in each group;
the analysis module III is used for combining all initial relevant risk factors acquired by the analysis module I and newly increased risk factors acquired by the analysis module II, bringing the combined initial relevant risk factors and newly increased risk factors into a multi-factor logistic regression model, and screening out independent risk factors by adopting a stepwise regression method;
and the model establishing module is used for assigning the independent risk factors and establishing a prediction model based on the multi-factor logistic regression model of the analysis module III.
9. An aortic dissection patient post-operative death prediction model building terminal device, the terminal device comprising an interface end, wherein the terminal device further comprises a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, wherein the processor when executing the computer program implements the aortic dissection patient post-operative death prediction model building method according to any one of claims 1 to 7.
10. A computer-readable storage medium comprising an interface plug, wherein the computer-readable storage medium further comprises a stored computer program, wherein the computer program when executed controls a device on which the computer-readable storage medium is located to perform the method for model building of postoperative death prediction in aortic dissection patients according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210250016.2A CN114864097A (en) | 2022-03-14 | 2022-03-14 | Method and device for establishing aorta dissection patient postoperative death prediction model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210250016.2A CN114864097A (en) | 2022-03-14 | 2022-03-14 | Method and device for establishing aorta dissection patient postoperative death prediction model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114864097A true CN114864097A (en) | 2022-08-05 |
Family
ID=82627809
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210250016.2A Pending CN114864097A (en) | 2022-03-14 | 2022-03-14 | Method and device for establishing aorta dissection patient postoperative death prediction model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114864097A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117153377A (en) * | 2023-10-11 | 2023-12-01 | 中山大学附属第一医院 | Model for predicting death risk of adult patient with moderately severe aortic valve stenosis |
-
2022
- 2022-03-14 CN CN202210250016.2A patent/CN114864097A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117153377A (en) * | 2023-10-11 | 2023-12-01 | 中山大学附属第一医院 | Model for predicting death risk of adult patient with moderately severe aortic valve stenosis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112837819B (en) | Method for establishing acute kidney injury prediction model after coronary artery bypass grafting operation | |
| US20190378619A1 (en) | Using machine learning to predict health conditions | |
| CN110246577B (en) | Method for assisting gestational diabetes genetic risk prediction based on artificial intelligence | |
| CN115019919B (en) | Deep learning-based method for establishing function prediction model after early recovery of stroke | |
| CN111312401B (en) | After-physical-examination chronic disease prognosis system based on multi-label learning | |
| CN106980757A (en) | The concurrent coronary artery pathological changes hazards management system of Kawasaki disease and method for digging | |
| CN111297329A (en) | Method and system for predicting dynamic morbidity risk of cardiovascular complications of diabetic patients | |
| Altikardes et al. | A novel method for dipper/non-dipper pattern classification in hypertensive and non-diabetic patients | |
| Pavithra et al. | Hybrid feature selection technique for prediction of cardiovascular diseases | |
| CN114864097A (en) | Method and device for establishing aorta dissection patient postoperative death prediction model | |
| CN114974585A (en) | Construction method of early risk prediction and evaluation model of metabolic syndrome in gestational period | |
| CN112820397B (en) | Method for establishing peri-operative risk prediction model of coronary artery bypass grafting | |
| Wadhawan et al. | ETCD: An effective machine learning based technique for cardiac disease prediction with optimal feature subset selection | |
| Hill et al. | Cardiogenic shock classification and associated mortality risk | |
| Segura et al. | Predicting cardiovascular disease by combining optimal feature selection methods with machine learning | |
| GÜLDOĞAN et al. | Performance evaluation of different artificial neural network models in the classification of type 2 diabetes mellitus | |
| CN116564512A (en) | Artificial intelligence prognosis evaluation system for heart failure and renal insufficiency patient | |
| Mendes et al. | Development and validation of a prediction score for low-cardiac-output syndrome after adult cardiac surgery | |
| Riyaz et al. | Ensemble Learning for Coronary Heart Disease Prediction | |
| Wang et al. | A local risk prediction model for prolonged ventilation after adult heart valve surgery in a Chinese single center | |
| CN111883258B (en) | Method for constructing OHSS indexing parting prediction model | |
| Xie et al. | A predictive model for postoperative adverse outcomes following surgical treatment of acute type A aortic dissection based on machine learning | |
| CN116525104B (en) | Quick staged diagnosis system, equipment and storable medium for cardiogenic shock | |
| CN114400094A (en) | Construction method and application of myocardial infarction risk prediction model after percutaneous coronary intervention treatment for multi-disease co-morbid old patients | |
| RU2759483C1 (en) | Method for predicting the deterioration of the condition of an elderly patient with myocardial infarction and senile asthenia syndrome |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |