A death risk prediction model for severe burn combined with liver enzyme abnormality, a construction method and a storage medium

Abstract

The present application relates to a death risk prediction model for severe burn combined with liver enzyme abnormality, a construction method and a storage medium. The construction method of the death risk prediction model comprises: collecting clinical feature data of patients with severe burn combined with liver enzyme abnormality; using an RF-RFE joint regression model to analyze and screen the clinical feature data of the training set and whether to die, and screening out the clinical feature data related to death of severe burn combined with liver enzyme abnormality; and based on the screened clinical feature data, a death risk prediction model is constructed. The present application also provides a death risk prediction model constructed by the construction method. The present application also provides a storage medium. Based on the clinical basic data and disease-related indicators of patients, the present application identifies the risk factors leading to the death of patients with severe burn and liver enzyme abnormality, so as to predict the death probability of patients in advance, conveniently, quickly and accurately, thereby helping doctors to take corresponding prevention and treatment measures for patients.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical diagnosis, in particular to a death risk prediction model for severe burn combined with liver enzyme abnormality, a construction method and a storage medium. BACKGROUND

[0002] Burn is a type of trauma that is extremely common in daily life, which poses a great threat to human life and health. Burn not only can lead to long-term physical and mental problems, but also causes a large number of patients to lose precious lives every year due to burn. The classification of burn mainly includes the depth of burn and the area of burn. The depth of burn ranges from first degree to fourth degree, and the area of burn is divided into two types of slight and severe. The feature of slight burn is that the area of burn is less than 10% of total body surface area (TBSA), mainly showing as superficial burn. In general, patients with total body surface area of burn > 20% or full-thickness burn > 10% are defined as severe burn.

[0003] Although the progress of resuscitation technology, antibiotic treatment, surgery and nutritional support has reduced the mortality rate of severe burn patients in recent years, their mortality rate is still high compared with other critically ill patients. One of the reasons for this phenomenon is the multiple organ dysfunction syndrome (MODS) caused by organ damage, especially for patients with total burn area (TBSA) reaching or exceeding 20%, the incidence of MODS can be as high as 63.4%. In MODS, liver damage is extremely common, although the research on the liver response to burn is relatively limited.

[0004] After burn, various factors can cause liver damage, including perfusion deficiency caused by hypotension, hepatocyte death induced by pro-inflammatory cytokines, liver edema and fat changes, etc. Studies have shown that the mechanism of liver damage may be related to ischemic hepatitis caused by effective blood volume reduction in the early stage of burn, and hepatocyte apoptosis and liver enzyme abnormalities caused by ischemia-reperfusion injury. Similarly, there is evidence that liver-related factors will affect the prognosis of burn patients, because the persistent disorder of liver enzymes indicates sepsis and poor prognosis of burn wounds. Common serum liver enzymes include aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP) and gamma glutamyl transferase (GGT), etc. Among them, AST and ALT are the most common liver enzymes, and they are usually measured to diagnose liver damage. The content of liver enzymes reflects the degree and type of liver damage.

[0005] Since the late 1930s, the occurrence of post-burn liver dysfunction has been reported. In the 1980s, a series of autopsies revealed the enlargement of the liver after burns. Subsequent studies have shown that the immediate increase in liver volume after burns is due to the formation of edema. The exacerbation of edema can cause cell damage and release liver enzymes, which return to normal levels during liver regeneration, thus affecting the prognosis. The traumatic response caused by severe burns is both general and specific, and is crucial for the epidemiological analysis and mortality risk assessment of patients with liver enzyme abnormalities. However, there is no prediction of whether a patient with severe burns combined with liver enzyme abnormalities will die in the future in the prior art, which leads to the inability to prevent and treat in time. Therefore, timely and accurate identification of risk factors for death is crucial for assessing the clinical prognosis of patients and guiding individualized treatment. SUMMARY

[0006] Therefore, the purpose of the present application is to provide a severe burn combined with liver enzyme abnormality death risk prediction model, a construction method and a storage medium, to identify the risk factors that cause the death of patients with severe burns combined with liver enzyme abnormalities based on the clinical basic data and disease-related indicators of patients, and to establish a well-validated death risk prediction model using these factors to predict the death probability of patients in advance, conveniently, quickly and accurately, thereby helping doctors to take appropriate prevention and treatment measures for patients.

[0007] In order to achieve the above-mentioned purpose, the technical scheme adopted by the present application is as follows:

[0008] A construction method of a severe burn combined with liver enzyme abnormality death risk prediction model, characterized in that it comprises the following steps:

[0009] S1, data collection and preprocessing: collecting and preprocessing the clinical feature data of patients with severe burns combined with liver enzyme abnormalities;

[0010] S2, feature data screening: dividing the preprocessed clinical feature data into a training set and a validation set, analyzing and screening the clinical feature data of the training set with a p value less than 0.05 and whether death occurs using an RF-RFE combined LASSO regression model, and screening out the clinical feature data related to death of severe burns combined with liver enzyme abnormalities;

[0011] S3, model construction: based on the screened clinical feature data related to death, a severe burn combined with liver enzyme abnormality death risk prediction model is constructed.

[0012] Among them, the liver enzyme abnormality is defined as at least one of the indicators of aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP) and gamma glutamyl transferase (GGT) in serum being higher or lower than the normal value.

[0013] Preferably, the clinical characteristic data comprises: age, gender, admission date, injury date, BMI, burn etiology, burn area, burn location, inhalation injury, length of stay, admission ALT, admission AST, degree of liver enzyme abnormality, duration, time of death, and length of hospital stay;

[0014] Preferably, the clinical characteristic data related to death of severe burn combined with liver enzyme abnormality comprises: age, total burn area (TBSA), and admission direct bilirubin.

[0015] Preferably, the death risk prediction model is a nomogram.

[0016] Preferably, in the nomogram:

[0017] The first column is a score scale (b(X-m) terms), with a score range of -2 to 9; the second column is admission direct bilirubin (DBIL), with a range of 0 to 280; the third column is total burn area (TBSA), with a range of 20 to 100; the fourth column is the age of the patient (age), with a range of 10 to 90; the fifth column is the total score, with a range of -4 to 10; and the sixth column is the probability of death risk, with a range of 0.01 to 0.998.

[0018] In the nomogram, the second to fourth rows are related factors, and different situations of different factors correspond to different scores of the scale; the fifth row is the sum of the scores of each factor, and the sixth row has a corresponding relationship with the total score of the fifth row. According to different scores, it is projected to the corresponding position, which is the corresponding death risk probability of severe burn combined with liver enzyme abnormality.

[0019] Preferably, the average AUROC value of the nomogram model is 0.806.

[0020] Preferably, the death risk prediction model further comprises a death risk probability calculation formula:

[0021] (I)

[0022] In formula (I), X1 represents age; X2 represents total burn area (TBSA); X3 represents admission direct bilirubin (DBIL); and P represents the death risk probability of severe burn combined with high creatinine abnormality.

[0023] Preferably, the p-value is obtained by t-test or chi-square test.

[0024] Preferably, it further comprises:

[0025] The clinical characteristic data is randomly divided into a 70% training set and a 30% validation set; the model is trained by the training data set to obtain a death risk prediction model;

[0026] verify the mortality risk prediction model by the verification set.

[0027] The application also provides a mortality risk prediction model of severe burn combined with liver enzyme abnormality, which is constructed by the construction method.

[0028] Preferably, the mortality risk prediction model of severe burn combined with liver enzyme abnormality comprises:

[0029] The data collection and preprocessing module is used for collecting clinical feature data of death patients and non-death patients of severe burn combined with liver enzyme abnormality, and pre-processing the clinical feature data.

[0030] The feature data screening module is used for dividing the pre-processed clinical feature data into a training set and a verification set, analyzing and screening the clinical feature data of the training set with a p value less than 0.05 and death by using a RF-RFE combined LASSO regression model, and screening the clinical feature data related to death of severe burn combined with liver enzyme abnormality.

[0031] The prediction model construction module is used for constructing the mortality risk prediction model of severe burn combined with liver enzyme abnormality based on the screened clinical feature data related to death.

[0032] The application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the steps of the construction method.

[0033] The application has the following beneficial effects:

[0034] The construction method of the mortality risk prediction model of severe burn combined with liver enzyme abnormality can ensure that the basic data for constructing the prediction model is comprehensive and representative by collecting the clinical feature data of patients with severe burn combined with liver enzyme abnormality; the data quality is improved by cleaning and standardizing the collected clinical feature data in the data preprocessing stage, and accurate and reliable data support is provided for subsequent analysis; the RF-RFE combined LASSO regression model is used for feature screening, which can effectively process high-dimensional data and screen clinical features with high correlation with death, thereby improving the prediction accuracy of the model; the model construction is not only based on the clinical manifestations of the clinical feature data, but also combined with experimental detection indexes, so that the prediction model is more comprehensive and can comprehensively reflect the severity of the patient's condition and the mortality risk; the mortality risk prediction model constructed based on the screened indexes can provide an accurate and practical tool for clinicians, realize the visualization of the results, help to evaluate the mortality risk of the patient, and thus develop a more personalized treatment plan, which has popularization and application value in the field of medical diagnosis technology. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 Flow chart of the process of collecting and preprocessing the clinical characteristics data of patients with severe burn combined with liver enzyme abnormalities;

[0036] Figure 2 Flow chart of RF-RFE combined LASSO for screening risk predictors of death, wherein 2(a) is RF-RFE based feature variable screening; 2(b) is LASSO regression (lambda: 1 SE) based feature variable screening; 2(c) is RF-RFE combined LASSO; LASSO: Least Absolute Shrinkage and Selection Operator; RFE: Recursive Feature Elimination RF: Random Forest;

[0037] Figure 3 Nomogram model of the death risk prediction model of patients with severe burn combined with liver enzyme abnormalities;

[0038] Figure 4 ROC curve of the validation set of the logistic regression model;

[0039] Figure 5 Calibration curve analysis result graph of the validation set of the nomogram model;

[0040] Figure 6 Calibration curve analysis result graph of the validation set of the nomogram model;

[0041] Figure 7 ROC curve of the training set of the logistic regression model;

[0042] Figure 8 Calibration curve analysis result graph of the training set of the nomogram model;

[0043] Figure 9 Calibration curve analysis result graph of the training set of the nomogram model. DETAILED DESCRIPTION

[0044] The present application will be described with reference to the attached drawings and preferred embodiments to thereby provide for a thorough understanding of the application to those skilled in the art. The application can be practiced with the claims hereinafter disclosed and can be practiced in various ways. There are numerous modifications and variations of the preferred embodiments as set forth herein that will be apparent to those skilled in the art, but will not necessarily fall within the scope of the application. Therefore, it is to be understood that the application is not to be limited to the specific embodiments disclosed and that modifications and / or

[0045] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present application in a schematic manner, and only the components related to the present application are shown in the diagrams, not the number, shape and size of the components when actually implemented. The actual implementation of each component can be a random change, and the component layout pattern can be more complex.

[0046] The present application aims to disclose a death risk prediction model for severe burn combined with liver enzyme abnormality, a construction method and a storage medium, to identify the risk factors leading to the death of patients with severe burn combined with liver enzyme abnormality based on the clinical basic data and disease-related indicators of the patients, and to establish a fully verified death risk prediction model by using these factors to predict the death probability of the patients in advance, conveniently, quickly and accurately, so as to help doctors take corresponding prevention and treatment measures for the patients.

[0047] The construction method of the death risk prediction model for severe burn combined with liver enzyme abnormality comprises the following steps:

[0048] S1, data collection: collecting the clinical characteristic data of the death patients and the non-death patients with severe burn combined with liver enzyme abnormality, and pre-processing the clinical characteristic data to remove the patients with incomplete clinical characteristic data;

[0049] S2, feature data screening: dividing the pre-processed clinical characteristic data into a training set and a validation set, and analyzing and screening the clinical characteristic data of the training set with a p value less than 0.05 in t test or chi-square test and whether death through RF-RFE combined LASSO regression model, screening the clinical characteristic data that meets the RF-RFE and LASSO regression screening results at the same time, and obtaining the clinical characteristic data related to the death of severe burn combined with liver enzyme abnormality;

[0050] S3, model construction: based on the screened clinical characteristic data related to the death, a death risk prediction model for severe burn combined with liver enzyme abnormality is constructed.

[0051] The liver enzyme abnormality is defined as at least one index of aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP) and gamma glutamyl transferase (GGT) in serum being higher or lower than the normal value.

[0052] Based on the R value (R=(ALT / ULN) / (ALP / ULN)), the damage mode is evaluated by diagnostic classification, and the coincidence rate is 96%. R≥5 is hepatocyte injury type liver enzyme abnormality; 2

[0053] Among them, liver enzyme abnormalities are defined as serum ALT, AST, ALP, GGT being at least one higher or lower than the normal value. Mild liver enzyme elevation is at least one content of serum ALT, AST, ALP, GGT being 1-5 times ULN, moderate liver enzyme elevation is at least one content of serum ALT, AST, ALP, GGT being 5-15 times ULN, and severe liver enzyme elevation is at least one content of serum ALT, AST, ALP, GGT being greater than 15 times ULN.

[0054] In some embodiments, the clinical feature data includes: age, gender, admission date, injury date, BMI, burn etiology, burn area, burn location, inhalation injury, hospitalization days, admission ALT, admission AST, liver enzyme abnormality degree, duration, etc., time of death, and hospitalization time.

[0055] In some embodiments, the clinical feature data related to death in severe burn combined with liver enzyme abnormalities includes: age, total burn area (TBSA), and admission direct bilirubin value.

[0056] In some embodiments, the death risk prediction model is a nomogram model.

[0057] As shown in S1, data collection and preprocessing includes:

[0058] As shown in S1, data collection and preprocessing includes: Figure 1 As shown in S1, data collection and preprocessing includes:

[0059] As shown in S2, feature data screening includes:

[0060] The patients with liver enzyme abnormalities were further divided into two groups (survival group (N=394) and non-survival group (N=85). The variables consistent with RF-RFE and LASSO screening were selected. The analysis results show that the demographic factors include age, the experimental detection indexes include admission direct bilirubin value, and the burn degree factors include total burn area (TBSA).

[0061] Logistic analysis revealed that mixed injury (OR, 4.43; 95% CI, 2.70-7.27; P < 0.001), hepatocellular injury (OR, 49.75; 95% CI, 16.25-152.31; P < 0.001), and AST / ALT (OR, 1.61; 95% CI, 1.35-1.92; P < 0.001) were considered risk factors for moderate to severe liver enzymes during hospitalization.

[0062] Of these, 479 patients developed abnormal liver enzymes, and 85 of them died, for a mortality rate of 17.75%. As shown in Table 1, the age of patients who died was older than that of the survivors (52.00 (43.00, 67.00) vs 46.00 (35.00, 53.00), p < 0.01). Moreover, the severity of burns was significantly higher in the deceased patients, including total burn area (54.00 (33.00, 85.00) vs 37.00 (28.00, 52.00), p < 0.001), Baux score (118.00 (101.00, 141.00) vs 90.00 (76.00, 108.00), p < 0.001), and burn index (61.50 (34.00) vs 33.50 (22.50, 53.00), 92.00), p < 0.001). It is worth noting that the AST of patients who died at admission was significantly higher (67.40 (47.90, 128.10) vs 37.00 (28.00, 52.00), p < 0.01) and total bilirubin was significantly higher (25.07 (14.00, 43.67) vs 17.40 (12.51, 25.63), p < 0.001).

[0063] Table 1 Comparison of clinical characteristics between the survival group and the death group in patients with abnormal liver enzymes

[0064]

[0065]

[0066]

[0067] RF-RFE combined with LASSO is used to screen risk predictors of death, such as Figure 2 As shown in the figure, a total of three non-zero characteristic variables (including total burn area, age, and direct bilirubin) were screened as constructed predictor variables. SPSS 27.0 software was used for regression analysis, nomogram construction, and evaluation.

[0068] For example, S3, model construction, includes: based on the above-mentioned screening of clinical features related to death data, based on LASSO regression model, construct nomogram to predict mortality, as shown in Figure 3

[0069] Among them, in the nomogram model:

[0070] The first column is the score scale (b(X-m) terms), the score range is -2~9; the second column is the direct bilirubin (DBIL) on admission, the range is 0~280; the third column is the total burn area (TBSA), the range is 20~100; the fourth column is the age of the patient (age), the range is 10~90; the fifth column is the total score, the range is -2~3; the sixth column is the death risk probability, the range is 0.01~0.998.

[0071] In the nomogram, the second to fourth rows are related factors, and different situations of different factors correspond to different scores of the scale; the fifth row is the sum of the scores of each factor, and the sixth row has a corresponding relationship with the total score of the fifth row. According to different scores, it is projected to the corresponding position, that is, the corresponding death risk probability of severe burn combined with liver enzyme abnormality.

[0072] The average AUROC value of the nomogram model is 0.806, the calibration slope is 1.245, and the Brier score is 0.131. It shows good fitting.

[0073] In some embodiments, the death risk prediction model further comprises a death risk value formula:

[0074] The death risk probability calculation formula is:

[0075] (I)

[0076] In formula (I), X1 represents age; X2 represents total burn area (TBSA); X3 represents direct bilirubin (DBIL) on admission; P represents the death risk probability of severe burn combined with creatinine abnormality.

[0077] According to the type and degree of liver enzyme abnormality, the patients with liver enzyme abnormality were analyzed. The variables missing more than 30% were deleted, and the remaining variables missing ratio was less than 4%, so the mean value was inserted. After interpolation, all clinical feature data of the patients with liver enzyme abnormality were randomly divided into training set (70%) and validation set (30%), and the prediction model was fitted using logistic regression. The estimated training cohort was screened for risk factors using LASSO regression and RF-RFE, the model was fitted and a dynamic nomogram was constructed. Finally, the model was evaluated on the validation set. The area under the receiver operating characteristic curve (AUC), the calibration curve and the decision curve analysis were used to evaluate the model, and the results are shown in Figures 4 to 7 ​​

[0078] wherein, Figure 4 is a plot of the work characteristic curve of the validation set subject cohort for the logistic regression model; Figure 5 is the calibration curve analysis result of the validation set cohort for the nomogram model; Figure 6 is the calibration curve analysis result of the validation set cohort for the nomogram model; Figure 7 is a plot of the work characteristic curve of the training set subject cohort for the logistic regression model; Figure 8 is the calibration curve analysis result of the training set cohort for the nomogram model; Figure 9 is the calibration curve analysis result of the training set cohort for the nomogram model. The decision curve analysis shows that the prediction model has good clinical benefit.

[0079] In some embodiments, a death risk prediction model for severe burn combined with liver enzyme abnormality is also provided, which is constructed by the construction method in any of the above embodiments, and the construction method comprises the following steps:

[0080] A data collection and preprocessing module is configured to collect clinical feature data of death patients and non-death patients with severe burn combined with liver enzyme abnormality, and preprocess the clinical feature data;

[0081] A feature data screening module is configured to divide the preprocessed clinical feature data into a training set and a validation set, analyze and screen the clinical feature data of the training set with a p value less than 0.05 and death by using a RF-RFE combined LASSO regression model, and screen out clinical feature data related to death of severe burn combined with liver enzyme abnormality;

[0082] A prediction model construction module is configured to construct a death risk prediction model for severe burn combined with liver enzyme abnormality based on the screened clinical feature data related to death.

[0083] In some embodiments, a computer readable storage medium having a computer program stored thereon is also provided, and the computer program is executed by a processor to implement the steps of the construction method in any of the above embodiments.

[0084] In summary, the severe burn combined with liver enzyme abnormality death prediction nomogram model of the present application can comprehensively evaluate the clinical characteristic indexes and laboratory data of the patient, reduce the judgment errors caused by one-sided information. When the severe burn patient has liver enzyme abnormality, whether the severe burn patient combined with liver enzyme abnormality will die can be predicted through the nomogram model based on the clinical characteristic indexes and laboratory data characteristics. The model of the present application is simple, intuitive, easy to popularize and apply, and can effectively enable the clinician to accurately individualize the evaluation of the severe burn patient when receiving treatment, bring better survival benefit to the patient, has important value for the effective application of severe burn combined with liver enzyme abnormality prognosis and monitoring, and provides personalized and professional health management scheme for different stratification, has popularization and application value in the field of medical diagnosis technology.

[0085] The above examples are only preferred embodiments for fully illustrating the present application, and the protection scope of the present application is not limited thereto. Any equivalent replacement or transformation made by the skilled in the art on the basis of the present application is within the protection scope of the present application.

Claims

1. A method for constructing a mortality risk prediction model for severe burns combined with abnormal liver enzymes, characterized in that: The following steps are involved: S1. Data collection and preprocessing: Clinical characteristics data of patients with severe burns and abnormal liver enzymes were collected and preprocessed; S2. Feature Data Screening: The preprocessed clinical feature data were divided into a training set and a validation set. The RF-RFE combined with the LASSO regression model was used to analyze and screen the clinical feature data in the training set with a p-value less than 0.05 and the relationship between the clinical feature data and death. Clinical feature data associated with severe burns combined with abnormal liver enzymes and death were screened. S3. Model construction: Based on the screened clinical characteristic data related to death, a mortality risk prediction model for severe burns with abnormal liver enzymes was constructed; The death risk prediction model is a nomogram model; In the nomogram model: The first column is the score scale (b(Xm)terms), with a range of -2 to 9; the second column is the admission direct bilirubin (DBIL), with a range of 0 to 280; the third column is the total burn area (TBSA), with a range of 20 to 100; the fourth column is the patient's age (age), with a range of 10 to 90; the fifth column is the total score, with a range of -4 to 10; the sixth column is the probability of death risk, with a range of 0.01 to 0.998; The mean AUROC value of the nomogram model was 0.806; The death risk prediction model also includes a death risk probability calculation formula: In formula (I), X1 represents age; X2 represents total burn area (TBSA); X3 represents direct bilirubin (DBIL) upon admission; and P represents the mortality risk probability of severe burns combined with abnormal liver enzymes.

2. The method for constructing a mortality risk prediction model for severe burns complicated with abnormal liver enzymes according to claim 1, characterized in that: The clinical characteristics data include: age, sex, admission date, injury date, BMI, burn etiology, burn area, burn location, inhalation injury, length of hospital stay, admission ALT, admission AST, admission bilirubin, degree of liver enzyme abnormality, duration, death time, and length of hospital stay; And / or, clinical characteristic data related to death of severe burns with abnormal liver enzymes, including: age, total burned area (TBSA) and direct bilirubin on admission.

3. The method for constructing a mortality risk prediction model for severe burns complicated with abnormal liver enzymes according to claim 1, characterized in that: The p-value is obtained by t-test or chi-square test; And / or, the construction method further comprises: The clinical characteristic data were randomly divided into 70% training set and 30% validation set; The model is trained using the training data set to obtain a death risk prediction model; The death risk prediction model was validated using the validation set.

4. A mortality risk prediction model for severe burns complicated with abnormal liver enzymes constructed by the construction method according to any one of claims 1 to 3.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the construction method according to any one of claims 1 to 3 are implemented.

Images