Acute hypotension noninvasive simple risk scoring method based on multi-factor logistic regression

Abstract

The invention discloses an acute hypotension noninvasive simple risk scoring method based on multi-factor logistic regression. The method comprises the following steps: 1) data extraction; 2) data processing; 3) constructing a risk scoring method; 4) evaluating the effect of the scoring method; and 5) performing external verification on the scoring method. The method has the advantage that a simple and efficient acute hypotension risk scoring model is constructed. According to the method, on the basis of fully considering physiological features, the pertinence is high, the prediction accuracy is high, and a reliable quantitative basis can be provided for early recognition and intervention of acute hypotension. The method can be widely applied to pre-hospital first aid, ICU monitoring and other medical scenes, and can also be effectively applied to the fields of health management, personalized prevention and the like, so that the early warning capability and management efficiency of acute hypotension events are improved, and the risk of serious complications of patients is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of acute hypotension scoring, and in particular to a non-invasive and simple risk scoring method for acute hypotension based on multivariate logistic regression. Background Art

[0002] Acute hypotension is a common and severe symptom in critically ill patients. It is often triggered by trauma, cardiac failure, massive bleeding, or systemic infection, posing a direct threat to patients' lives. However, if acute hypotension is detected early and appropriate intervention is taken, the course of the disease can be reversed in some cases, significantly reducing mortality and improving patient outcomes.

[0003] Therefore, early detection of acute hypotension events and prediction of their likelihood or duration are crucial for developing effective interventions. This not only helps prevent worsening of the condition and reduce irreversible consequences such as shock and organ failure, but also significantly reduces mortality, playing a crucial role in improving the treatment of emergency and critically ill patients.

[0004] Although there are several scoring tools widely used in critical care and acute disease assessment, these scoring tools are mainly used to assess the severity of multiple organ failure and systemic diseases. Therefore, there is still no scoring tool specifically for acute hypotension, especially a tool that can quickly and accurately assess risk in emergency and critical care. Existing scoring systems generally have the following problems and limitations: Complexity and low timeliness: Existing scoring tools such as the SOFA score usually rely on multiple clinical data items, and the calculation process is cumbersome and takes a long time to complete. This makes these tools unsuitable for early warning of acute hypotension and unable to provide immediate and actionable intervention guidance.

[0005] Poor universality: Although existing scoring methods have certain applicability in specific scenarios such as intensive care, they are not optimized for the specific condition of acute hypotension, and their scope of application is limited, making it difficult to popularize them in pre-hospital emergency care or routine clinical monitoring.

[0006] High cost and invasiveness: Many traditional scoring methods rely on expensive tests (such as blood biochemistry, blood routine, coagulation function, etc.), and some require invasive operations, which makes them difficult to use in emergency and non-professional environments. Summary of the Invention

[0007] (1) Technical problems solved The goal of this invention is to develop an acute hypotension risk scoring tool based on seven core indicators. This tool, which enables noninvasive, real-time monitoring via wearable devices, addresses the timeliness, simplicity, and affordability shortcomings of existing scoring systems. This tool can provide rapid and accurate risk assessments, particularly in the early stages of acute hypotension, enabling more timely early warning and intervention strategies.

[0008] Currently, with the rapid development of big data technology and intelligent medical devices, scoring tools based on data modeling and analysis are becoming increasingly popular. Multivariate logistic regression models have been widely used in various risk assessment systems. For example, the mortality risk scoring tool based on logistic regression can categorize patient risk into multiple levels and provide data support for clinical decision-making. However, despite the existence of a number of scoring systems, a simple and efficient scoring tool specifically for acute hypotension has not yet emerged. In particular, real-time monitoring tools that can be integrated with wearable devices are still under development.

[0009] This study aims to develop a simplified, clinically applicable risk scoring tool using multivariate logistic regression combined with the clinical characteristics of acute hypotension. This tool not only dynamically assesses the risk of acute hypotension but also enables noninvasive, rapid, and real-time monitoring through smart devices, offering significant application value, particularly in critical areas such as pre-hospital emergency care, ICU monitoring, and prognosis assessment.

[0010] In summary, the present invention aims to fill the gap in the existing technology and provide a more convenient, accurate and economical acute hypotension risk assessment tool to quickly identify the risk of acute hypotension and assist medical staff in making more accurate clinical decisions in emergency and critical care treatment, thereby improving patient prognosis and improving treatment efficiency.

[0011] (2) Technical solution To achieve the above objectives, the present invention provides the following technical solution: a simple non-invasive risk scoring method for acute hypotension based on multivariate logistic regression, characterized in that it comprises the following steps: 1) Data extraction: Based on the inclusion and exclusion criteria, seven indicators of patients with acute hypotension after infection were extracted from the database and the outcome markers were calculated. The incidence of in-hospital acute hypotension was marked with 1 for occurrence and 0 for non-occurrence. Acute hypotension patients and non-acute hypotension patients were identified. Seven indicators (systolic blood pressure, diastolic blood pressure, respiratory rate, heart rate, body temperature, blood oxygen saturation, and age) of patients with acute hypotension and non-acute hypotension were extracted from the database as risk score indicators. 2) Data processing: clean the indicator data to deal with outliers, fill in the gaps, add missing values, and unify the units; 3) Constructing a risk scoring method, (1) Seven risk score indicators are included in the multivariate logistic regression model, and the regression coefficients of each risk score indicator are , the intercept ; (2) According to the clinical significance, the possible value range of each risk score indicator is grouped, and a suitable value is selected as the reference value of the group in each group. Generally, the middle value is selected as the reference value, denoted as ; (3) For each risk score indicator, a suitable group is selected as the basic risk group, and the reference value of the group is the basic risk reference value , which is identified in clinical practice as the closest indicator data range to non-acute hypotension patients; (4) Combine the regression coefficients obtained by the multivariate logistic regression model and the reference values of each group of risk score indicators , calculate the relative distance between the reference value of each group of risk score indicators and the basic risk reference value , the calculation formula is as follows, (5) Set an appropriate constant B value, which determines the degree of risk change when the score increases by one point, (6) Calculate the score corresponding to each group of risk score indicators , round the calculated value to the nearest integer, which is the score corresponding to the group. The calculation formula is as follows, (7) According to the results of step (6), add up the scores of each risk score indicator to calculate the total score P, the calculation formula is as follows (2-3); then according to the formula of the multivariate logistic regression model: (2-4) and (2-5), calculate the risk prediction probability value corresponding to each score , where the new coefficient B in the formula of the multivariate logistic regression model is determined in step (5), and the new intercept The calculation formula is shown in equation (2-6).

[0012] (2-3) (2-4) (2-5) (2-6) 4) Effect evaluation of the scoring method In order to evaluate the effect of the scoring method, the accuracy, recall rate, F1 score and area under the curve (AUC) are selected as evaluation indexes, and the definitions of the evaluation indexes are as follows, In the sample, TP, TN, FP and FN represent: TP: the number of samples of acute hypotension correctly identified as acute hypotension; TN: the number of samples without acute hypotension correctly identified as non-acute hypotension; FP: the number of samples without acute hypotension incorrectly identified as acute hypotension; FN: the number of samples of acute hypotension incorrectly identified as non-acute hypotension; Accuracy is the percentage of correctly predicted results in total samples, and its expression is: Recall rate is the probability of being predicted as a positive sample in the actual positive sample, and its expression is: F1 score considers both precision and recall, and balances them to achieve the highest, and its expression is:

[0013] The area under the curve (AUC) score is used to evaluate the prediction performance of the model; in order to calculate the AUC score, the true positive rate and false positive rate curve, i.e. ROC curve, is drawn, and then the area under the curve (AUC) is calculated to distinguish the two diagnostic groups of acute hypotension and non-acute hypotension, and the specific formula is as follows: ; 5) External validation of the scoring method, in order to further evaluate the application effect of the scoring method, the scoring method is externally validated.

[0014] Compared with the prior art, the present application has obvious advantages and beneficial effects in the risk assessment of acute hypotension, which are embodied in the following aspects: Timeliness: The suddenness and high risk of acute hypotension require effective risk assessment and intervention within a short period of time. Unlike traditional risk scoring methods that rely on complex clinical tests, the present application uses seven core indicators as evaluation indicators, which can be monitored in real time and continuously by wearable devices and monitors, ensuring that early warning information can be provided quickly at the critical moment of acute hypotension, greatly improving the timeliness of emergency response.

[0015] Economical: Compared with blood biochemical, blood routine, coagulation and other laboratory tests, the collection cost of the indicators is lower, and there is no need for complex laboratory operation and equipment support. By selecting seven basic indicators as evaluation indicators, the invention not only reduces the implementation cost, but also improves the technical popularization, so that the risk assessment of acute hypotension can be applied in more scenarios, including pre-hospital emergency and primary medical institutions.

[0016] Convenience: The invention uses wearable devices for data collection, with the characteristics of non-invasive, simple and fast operation. Patients do not need to perform invasive operations, but can monitor indicators in real time through wearable devices, which is convenient for wide application in emergency, ICU monitoring, health management and other scenarios. This simple operation method greatly reduces the technical requirements for medical personnel, improves the convenience of clinical operation, and is especially suitable for non-professionals and emergency personnel in emergency situations.

[0017] Accuracy: The scoring method of the invention focuses on seven core indicators, avoiding the complex and redundant data input process in traditional scoring methods, and can obtain accurate risk assessment results in a short time through simple indicator data. This efficiency not only improves the accuracy of early identification of acute hypotension, but also effectively reduces the possibility of misdiagnosis or missed diagnosis, ensuring that patients can receive timely intervention and treatment.

[0018] Wide application: This technical solution is not only suitable for pre-hospital emergency, intensive care and other high-risk clinical environments, but also can be applied to health management and personalized prevention fields combined with wearable devices. This provides a new idea for the prevention and treatment of acute hypotension, with great popularization value and application potential.

[0019] In summary, the invention simplifies the indicators, improves the real-time monitoring and economy, and enhances the convenience of operation, making the risk assessment of acute hypotension more efficient, accurate and easy to operate, providing a practical solution for early warning and intervention of acute hypotension.

[0020] Preferably, in step 1, acute hypotension patients and non-acute hypotension patients are determined from the Chinese People's Liberation Army General Hospital Emergency Database and MIMIC-IV Database.

[0021] Preferably, in step 2, multiple imputation method is used to fill in the data. In order to avoid the reduction of statistical test efficiency and bias caused by directly excluding missing values, multiple imputation method based on random forest is used to estimate the missing values in the experimental data, and H-L test is used.

[0022] Adopting the above scheme: 1. The application only selects seven core indicators as risk score indicators, accurately quantifies the risk of acute hypotension events, and constructs a simple and efficient acute hypotension risk score model. This method has high prediction accuracy and is highly targeted based on full consideration of physiological characteristics, and can provide reliable quantitative basis for early identification and intervention of acute hypotension.

[0023] 2. The scoring method of the application has wide applicability and can be combined with existing wearable devices to provide real-time acute hypotension risk monitoring function. This technology can not only be widely used in pre-hospital emergency, ICU monitoring and other medical scenes, but also can be effectively applied in health management, personalized prevention and other fields, thereby improving the early warning ability and management efficiency of acute hypotension events and reducing the risk of serious complications in patients.

[0024] 3. The application proposes a complete and rigorous scoring method construction process, covering data extraction, cleaning, processing from the first aid database, to the establishment of the final risk score model. Each link is designed and verified in detail to ensure the accuracy of the data and the scientificity of the scoring results. This process not only guarantees the efficiency and operability of the model, but also provides solid data support and theoretical basis for various emergency and monitoring situations in practical application.

[0025] (Three) beneficial effects Compared with the prior art, the application provides a multi-factor logistic regression-based acute hypotension non-invasive simple risk scoring method, which has the following beneficial effects: 1. The multi-factor logistic regression-based acute hypotension non-invasive simple risk scoring method only selects seven core indicators as risk score indicators, accurately quantifies the risk of acute hypotension events, and constructs a simple and efficient acute hypotension risk score model. This method has high prediction accuracy and is highly targeted based on full consideration of physiological characteristics, and can provide reliable quantitative basis for early identification and intervention of acute hypotension.

[0026] 2. The multi-factor logistic regression-based acute hypotension non-invasive simple risk scoring method can be combined with existing wearable devices to provide real-time acute hypotension risk monitoring function. This technology can not only be widely used in pre-hospital emergency, ICU monitoring and other medical scenes, but also can be effectively applied in health management, personalized prevention and other fields, thereby improving the early warning ability and management efficiency of acute hypotension events and reducing the risk of serious complications in patients.

[0027] 3、A complete and rigorous scoring method construction process is proposed, covering data extraction, cleaning, processing from the emergency database, to the establishment of the final risk scoring model. Each link is designed and verified in detail to ensure the accuracy of the data and the scientificity of the scoring results. This process not only guarantees the efficiency and operability of the model, but also provides solid data support and theoretical basis for various emergency and monitoring situations in practical application. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 The flow chart of the acute hypotension non-invasive simple risk scoring method based on multi-factor logistic regression proposed in the application; Figure 2 The data inclusion flow chart of the acute hypotension non-invasive simple risk scoring method based on multi-factor logistic regression proposed in the application; Figure 3 The scoring method construction table of the acute hypotension non-invasive simple risk scoring method based on multi-factor logistic regression proposed in the application; Figure 4 The acute hypotension risk scoring and probability table; Figure 5 The ROC curve chart; Figure 6 The external verification data inclusion flow chart in the application; Figure 7 The ROC curve of 4 scoring methods; Figure 8 The DCA curve of 4 scoring methods; Figure 9 The clinical commonly used scoring method. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0030] Please refer to Figures 1-9 , the acute hypotension non-invasive simple risk scoring method based on multi-factor logistic regression, comprising the following steps: 1) Data extraction, according to the inclusion and exclusion criteria, seven indicators of patients with acute hypotension after infection are extracted from the database, and the outcome identifier is calculated, indicating the occurrence of acute hypotension in hospital, 1 indicating occurrence, 0 indicating non-occurrence, and determining acute hypotension patients and non-acute hypotension patients. Seven indicators (systolic blood pressure, diastolic blood pressure, respiratory rate, heart rate, body temperature, oxygen saturation, age) of acute hypotension patients and non-acute hypotension patients are extracted from the database as risk score indicators; 2) Data processing, data cleaning is performed on the indicator data to process outliers, and data is supplemented to supplement missing values and unify units; 3) Construction of risk score method, (1) Seven risk score indicators are included in the multivariate logistic regression model, and the regression coefficient of each risk score indicator is , and the intercept is ; (2) According to the clinical significance, the possible value range of each risk score indicator is grouped, and a suitable value is selected as the reference value of the group in each group. Generally, the middle value is selected as the reference value, denoted as ; (3) For each risk score indicator, select a suitable group as the basic risk group, and the reference value of this group is the basic risk reference value , which is recognized in clinical practice as the closest indicator data range to non-acute hypotension patients; (4) Combine the regression coefficient obtained by the multivariate logistic regression model and the reference value of each group of risk score indicators , calculate the relative distance between the reference value of each group of risk score indicators and the basic risk reference value , and the calculation formula is as shown in the following formula, (5) Set an appropriate constant B value, which determines the degree of risk change when the score increases by one point, (6) Calculate the score corresponding to each group of risk score indicators , round the calculated value to the nearest integer, which is the score corresponding to the group. The calculation formula is as shown in the following formula, (7) According to the results of step (6), add up the scores of each risk score indicator to calculate the total score P, and the calculation formula is as follows (2-3); Then, according to the formulas of the multivariate logistic regression model: (2-4) and (2-5), the risk prediction probability value corresponding to each score is calculated , wherein the new coefficient B in the formula of the multivariate logistic regression model is determined in step (5), and the new intercept The calculation formula is shown in the following formula (2-6).

[0031] (2-3) (2-4) (2-5) (2-6) 4) Effect evaluation of the scoring method In order to evaluate the effect of the scoring method, the accuracy, recall rate, F1 score and area under the curve (AUC) are selected as evaluation indexes, and the definitions of the evaluation indexes are as follows, In the sample, TP, TN, FP and FN represent: TP: the number of samples of acute hypotension correctly identified as acute hypotension; TN: the number of samples without acute hypotension correctly identified as non-acute hypotension; FP: the number of samples without acute hypotension incorrectly identified as acute hypotension; FN: the number of samples of acute hypotension incorrectly identified as non-acute hypotension; The accuracy is the percentage of correctly predicted results in the total samples, and its expression is: The recall rate is the probability of being predicted as a positive sample in the actual positive sample, and its expression is: The F1 score considers the accuracy and recall rate, and balances them to the highest, and its expression is:

[0032] The area under the curve (AUC) score index is used to evaluate the prediction performance of the model; in order to calculate the AUC score, the true positive rate and false positive rate curve, i.e. the ROC curve, is drawn, and then the area under the curve (AUC) is calculated to distinguish the two diagnostic groups of acute hypotension and non-acute hypotension, and the specific formula is as follows: ; 5) External verification of the scoring method, in order to further evaluate the application effect of the scoring method, the scoring method is externally verified.

[0033] In step 1, acute hypotension patients and non-acute hypotension patients were identified from the PLAGH-ERD and MIMIC-IV databases. In step 2, multiple imputation was used to fill in the missing data. To avoid the reduction of statistical test power and bias caused by directly excluding missing values, a multiple imputation method based on random forest was used to estimate the missing values in the experimental data, and the H-L test was used.

[0034] When the method is implemented: 1. Data extraction According to the inclusion and exclusion criteria, seven index data of patients with acute hypotension after infection were extracted from the database, and the outcome identifier was calculated, which was marked as the occurrence of acute hypotension in hospital (1 represents occurrence, 0 represents non-occurrence).

[0035] Specifically, the data in the database comes from the PLAGH-ERD. It records in detail the diagnosis and treatment information of all emergency, critical and severe patients treated by the rescue unit of the hospital from 2015 to 2022, including diagnosis, medical orders, indicators, laboratory tests, imaging tests, etc. Among them, the indicator data is the most complete, with a collection frequency of about every 30 minutes.

[0036] 2、Data processing The indicator data is cleaned to handle outliers and fill in missing values.

[0037] In this step, multiple imputation is used to fill in the missing data. To avoid the reduction of statistical test power and bias caused by directly excluding missing values, a multiple imputation method based on random forest is used to estimate the missing values in the experimental data, and the H-L test is used.

[0038] 3、Construction of scoring method (1) The seven risk score indicators (systolic blood pressure, diastolic blood pressure, heart rate, respiratory rate, body temperature, oxygen saturation, and age) are included in the multivariate logistic regression model, and the regression coefficients of each risk score indicator are 0.0890, -0.0237, 0.0335, 0.0418, 0.2160, 0.0098, and 0.0270, respectively. The intercept is -1.6314-0.5032.

[0039] (2) According to the clinical significance, the possible value range of the seven risk score indicators is grouped, and the middle value in each group is selected as the reference value. The reference value of the last group of each risk score indicator is selected as the 0.5 quantile value of the data within the range.

[0040] (3) For each risk score indicator, select a suitable group as the basic risk group, and the reference value of this group is the basic risk reference value. The corresponding score of this group is 0.

[0041] (4) Calculate the relative distance of each group of risk score indicators to its base risk reference value, and the results are shown in Table 4. Figure 3

[0042] (5) Set the constant B to 10 .

[0043] (6) Calculate the score corresponding to each group of risk score indicators. The calculated value is rounded to the nearest integer, which is the score corresponding to the group, as shown in Table 5. Figure 4

[0044] (7) Add the scores of each risk score indicator to calculate the total score. According to the formula of the multi-factor logistic regression model, calculate the risk prediction probability value corresponding to each score, 4. Effect evaluation of scoring method In order to evaluate the effect of the scoring method, accuracy, recall, F1 score, and area under the curve (AUC) are selected as evaluation indicators. The definitions of the evaluation indicators are as follows.

[0045] In the sample, TP, TN, FP, and FN represent: TP: The number of samples of acute hypotension correctly identified as acute hypotension; TN: The number of samples without acute hypotension correctly identified as non-acute hypotension; FP: The number of samples without acute hypotension incorrectly identified as acute hypotension; FN: The number of samples of acute hypotension incorrectly identified as non-acute hypotension.

[0046] Accuracy (Accuracy) is the percentage of correctly predicted results in the total sample, and its expression is: (7-1) Recall (Recall) is the probability of being predicted as a positive sample among the actual positive samples, and its expression is: (7-2) F1 score considers both precision and recall, and balances them to achieve the highest, and its expression is: (7-3) Area under the curve (AUC) score is used to evaluate the prediction performance of the model. In order to calculate the AUC score, the true positive rate (TPR) and false positive rate (FPR) curve, i.e. ROC curve, needs to be drawn, as shown in Figure 1. Figure 5 ​​The two diagnostic groups of acute hypotension and non-acute hypotension were then distinguished by calculating the area under the curve (AUC), with the formula as follows: (7-4) (7-5) The maximum value of the AUC score is 1, which means that the classifier model can perfectly distinguish all samples. Therefore, the larger the AUC value, the better the prediction performance of the model.

[0047] The constructed scoring method was evaluated, with an accuracy of 0.80, a recall of 0.78, an F1 score of 0.72, and an AUC value of 0.86 (95% confidence interval of 0.853 to 0.873). The results show that the model has excellent ability to distinguish between positive and negative classes, and the prediction probability can truly reflect the actual risk.

[0048] 5. External validation of the scoring method To further evaluate the application effect of the scoring method, external validation of the scoring method was performed.

[0049] In external validation, the Medical Information Mart for Intensive Care (MIMIC)-IV database was used. Compared with MIMIC-III, MIMIC-IV has made many improvements, including data updates and reconstruction of some tables. This database collects clinical data of more than 190,000 patients and 450,000 hospitalization records received by Beth Israel Deaconess Medical Center (BIDMC) from 2008 to 2019. The database records detailed information of patients, including demographic information, laboratory test results, medication, indicators, surgical operations, disease diagnosis, drug management, and follow-up survival status.

[0050] The external validation results show that the accuracy of the scoring method is 0.82, the recall is 0.98, the F1 score is 0.77, and the AUC value is 0.96 (95% confidence interval of 0.951 to 0.968). These results indicate that the model has good effect and is well accepted in clinical medicine.

[0051] 6. Comparison of the scoring method with other methods To further compare the performance of the acute hypotension scoring method with other existing methods, the SOFA score, SIRS score, and SI score of these samples were extracted or calculated from the MIMIC IV database to predict the onset of acute hypotension, and the ROC curves and decision curve analysis curves of the four methods for predicting the onset of acute hypotension were compared by drawing them.

[0052] Decision Curve Analysis (DCA) is a method for evaluating the value of a predictive model in actual clinical decision-making. It achieves this by comparing the net benefits of different decision-making strategies within a specific threshold range. The "net benefit" here refers to the net effect after considering the benefits and losses brought about by false positive and false negative results. The core principle of DCA is to quantify the utility of a predictive model at a specific clinical decision threshold and compare it with other decision strategies, such as treating all patients or not treating any patients.

[0053] N: sample size; T: Threshold Value, which refers to the point at which a clinical decision maker (such as a doctor) makes a medical decision when the risk probability value is large; net benefit: represents the benefits of intervention when positive minus the losses (which need to be multiplied by the threshold value) when negative, the calculation formula is as follows: (7-6) The curve formed by the combination of "benefit values" at different 'thresholds' is called the DCA curve. Clinicians need to make decisions at the most suitable 'threshold'. Therefore, the DCA curve can help clinicians make decisions with the optimal 'threshold'.

[0054] As shown in Figure 7 , the AUC value of the acute hypotension scoring method is 0.96, while the AUC value of the SOFA score is 0.78, the SIRS score is 0.44, and the SI score is 0.84. In the DCA curve ( Figure 8 ), the threshold value ranges from 0 to 1, and the benefit of the acute hypotension scoring method is always the highest. When the threshold value is between 0.1 and 0.2, the benefit of the SOFA score is higher than that of the SI score, and the benefit of the SIRS score is the lowest. The results prove that the acute hypotension scoring method performs best.

[0055] 7. Score tool card diagram Comparison of alternatives: Although there is no existing scoring method specifically for acute hypotension, there are some widely used scoring systems for critical care and acute illness assessment in clinical practice. For example, Sequential Organ Failure Assessment (SOFA), Systemic Inflammatory Response Syndrome (SIRS), Oxford Acute Severity Score (OASIS), Shock Index (SI), etc. as shown in Figure 9 , have been widely used in the monitoring and prognosis of critically ill patients.

[0056] These scoring methods usually rely on multiple physiological parameters and are mainly used to assess organ failure, disease severity and prognosis of critically ill patients. However, these scoring systems are usually complex, cover a large number of indicators, and require detailed clinical data input, so they have certain limitations for early identification and intervention of acute hypotension.

[0057] The present application simplifies the evaluation process by selecting seven core indicators as risk scoring indicators, making the scoring method not only more efficient and easy to operate, but also particularly suitable for early screening of acute hypotension. Compared with existing scoring systems, the method of the present application has obvious advantages in targeting and simplicity, especially in pre-hospital emergency and wearable device application scenarios, it can provide more timely and accurate acute hypotension risk assessment.

[0058] In addition, although existing scoring methods can provide certain reference for clinical judgment, they usually cannot monitor acute hypotension risk in real time. Therefore, the present application, combined with the application of wearable devices, can realize dynamic and real-time monitoring of acute hypotension risk, providing more accurate and timely support for clinical practice.

[0059] In summary, 1. The present application only selects seven core indicators as risk scoring indicators, and by accurately quantifying the risk of acute hypotension events, a simple and efficient acute hypotension risk scoring model is constructed. This method is highly targeted and has high prediction accuracy based on consideration of physiological characteristics, and can provide reliable quantitative basis for early identification and intervention of acute hypotension.

[0060] 2. The scoring method of the present application has wide applicability and can be combined with existing wearable devices to provide real-time acute hypotension risk monitoring function. This technology can be widely used in pre-hospital emergency, ICU monitoring and other medical scenarios, and can also be effectively applied in health management, personalized prevention and other fields, thereby improving the early warning ability and management efficiency of acute hypotension events and reducing the risk of patients developing serious complications.

[0061] 3、The application proposes a complete and rigorous scoring method construction process, covering data extraction, cleaning, processing from the first aid database, to the establishment of the final risk scoring model. Each link is designed and verified in detail to ensure the accuracy of the data and the scientificity of the scoring results. This process not only guarantees the efficiency and operability of the model, but also provides solid data support and theoretical basis for various first aid and monitoring situations in practical application.

[0062] It should be noted that, in this paper, relational terms such as first and second are used merely to distinguish one entity or action from another, without necessarily requiring or implying that there is any such actual relationship or order between or among the entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without more limitations, an element defined by the statement "comprises one" does not exclude the existence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0063] Although embodiments of the application have been shown and described, it is to be understood that various modifications, substitutions, replacements and variations can be made to these embodiments without departing from the principles and spirit of the application, and the scope of the application is defined by the appended claims and their equivalents.

Claims

1. A simple non-invasive risk scoring method for acute hypotension based on multivariate logistic regression, characterized by: The following steps are involved: 1) Data extraction: According to the inclusion and exclusion criteria, seven indicators of patients with acute hypotension after infection were extracted from the database and the outcome markers were calculated. The incidence of acute hypotension in the hospital was marked with 1 for occurrence and 0 for non-occurrence. The patients with acute hypotension and those without acute hypotension were identified. The seven indicators of patients with acute hypotension and those without acute hypotension were extracted from the database as risk score indicators. 2) Data processing: clean the indicator data to deal with outliers, fill in the gaps, add missing values, and unify the units; 3) Constructing a risk scoring method, (1) The seven risk scoring indicators were incorporated into the multivariate logistic regression model, and the regression coefficients of each risk scoring indicator were obtained as follows: ,intercept ; (2) Group the possible value ranges of each risk score indicator according to clinical significance, and select an appropriate value in each group as the reference value of the group. Generally, the middle value is selected as the reference value, which is recorded as ; (3) For each risk scoring indicator, select an appropriate group as the basic risk group, and the reference value of this group is the basic risk reference value This grouping is clinically identified as being closest to the index data range of patients with non-acute hypotension; (4) Regression coefficients obtained by combining the multivariate logistic regression model and reference values ​​for each group of risk score indicators Calculate the relative distance between the reference value of each group of each risk score indicator and its basic risk reference value , the calculation formula is as follows: (5) Set an appropriate constant B value, which determines the degree of risk change when the score increases by one point. (6) Calculate the scores corresponding to each group of risk scoring indicators , round the calculated value to the nearest integer, which is the score corresponding to the group. The calculation formula is as follows: (7) Based on the results of step (6), the scores of each risk score indicator are added together to calculate the total score P, which is calculated using the following formula (2-3); then, according to the formulas (2-4) and (2-5) of the multivariate logistic regression model, the risk prediction probability value corresponding to each score is calculated. , where the new coefficient B in the formula of the multivariate logistic regression model is determined in step (5), and the new intercept The calculation formula is shown in the following formula (2-6); （2-3） （2-4） （2-5） （2-6） 4) Evaluation of the effectiveness of the scoring method In order to evaluate the effectiveness of the scoring method, accuracy, recall, F1 score, and area under the curve (AUC) were selected as evaluation indicators. The definitions of the evaluation indicators are as follows: In the sample, TP, TN, FP, and FN stand for: TP: the number of samples with acute hypotension correctly identified as acute hypotension; TN: number of samples without acute hypotension correctly identified as non-acute hypotension; FP: number of samples without acute hypotension that were incorrectly identified as having acute hypotension; FN: number of samples with acute hypotension incorrectly identified as non-acute hypotensive; The accuracy rate is the percentage of correct prediction results in the total samples, and its expression is: The recall rate is the probability of being predicted as a positive sample in the actual positive sample, and its expression is: The F1 score takes into account both precision and recall, maximizing both and achieving a balance. Its expression is: The area under the curve (AUC) score is used to evaluate the predictive performance of the model. To calculate the AUC score, it is necessary to plot the true positive rate and false positive rate curves, namely the ROC curve, and then calculate the area under the curve (AUC) to distinguish between the two diagnostic groups of acute hypotension and non-acute hypotension. The specific formula is as follows: ； 5) External validation of the scoring method. In order to further verify the application effect of the scoring method, the scoring method is externally validated.

2. The non-invasive simple risk scoring method for acute hypotension based on multivariate logistic regression according to claim 1, characterized in that: In step 1, acute hypotension patients and non-acute hypotension patients are determined from the database.

3. The non-invasive simple risk scoring method for acute hypotension based on multivariate logistic regression according to claim 1, characterized in that: In step 2, the data are filled with multiple interpolation methods. In order to avoid the reduction of statistical test efficiency and bias caused by directly excluding missing values, the multiple interpolation method based on random forest is used to estimate the missing values ​​in the experimental data and pass the HL test.

Images