A method and system for risk assessment of high altitude disease based on multivariate logistic regression

Abstract

This invention provides a method and system for assessing the risk of high-altitude diseases based on multivariate logistic regression, belonging to the field of high-altitude disease risk assessment technology. The method includes: collecting multidimensional characteristic data of individuals, including demographics, high-altitude exposure-related data, clinical symptoms, vital signs, and laboratory test indicators; inputting a multivariate logistic regression model containing risk sub-models for mild acute high-altitude reaction and severe high-altitude pulmonary edema. During model training, synthetic minority class oversampling is used to handle class imbalance, and stepwise regression combined with clinical priors is used to select independent variables; first and second risk probability values ​​are output separately, and after calibration according to scenario parameters, the risk levels for mild and severe illness are obtained by comparing with a stratified threshold set, generating an integrated assessment result covering the entire cycle from early onset to severe disease progression. This solves the problems of existing models having limited disease spectrum coverage, incomplete feature systems, poor interpretability, and a lack of scenario-based, full-cycle risk stratification capabilities.

Description

Technical Field

[0001] This invention relates to the field of altitude sickness risk assessment technology, and more specifically, to a method and system for altitude sickness risk assessment based on multivariate logistic regression. Background Technology

[0002] The hypoxic environment of high altitudes is the core trigger for acute mountain sickness (APS) in people rapidly ascending to altitudes above 2500m. As an idiopathic disease of high altitude, APS is characterized by mild symptoms such as acute mountain sickness (AMS), including headache, fatigue, and dizziness. Severe cases can progress to life-threatening complications such as high-altitude pulmonary edema (HAPE) and high-altitude cerebral edema, exhibiting rapid onset, rapid progression, and high mortality risk. Statistics show that over 100 million people from low-altitude areas worldwide enter high-altitude regions annually for tourism, work, military activities, etc., with an overall APS incidence rate of 30%-70%, including approximately 1% of severe HAPE. Failure to identify and intervene promptly poses a serious threat to public health. Therefore, developing accurate, interpretable, and easily scalable risk assessment methods for altitude sickness, enabling early screening and risk stratification, is a core requirement in clinical practice and public health prevention and control in high-altitude medicine. With the development of electronic health data, clinical testing technology, and machine learning algorithms, researchers both domestically and internationally have conducted a series of studies on the prediction of altitude sickness risk. Existing technologies include schemes that integrate demographic, vital sign, and biochemical indicators based on ensemble learning algorithms and resampling techniques to construct risk prediction models for high-altitude pulmonary edema; schemes targeting specific high-altitude work populations that employ random forests and logistic regression algorithms to construct altitude sickness risk prediction models, enabling preliminary predictions of disease risk among workers; some studies have used univariate and multivariate logistic regression analysis to screen key predictive variables for high-altitude pulmonary edema, combining this with nomograms to achieve self-assessment of disease risk; and other studies have focused on early warning of acute altitude sickness, constructing warning models based on thyroid function-related physiological indicators, providing a technical pathway for predicting the risk of acute altitude sickness. These existing studies have promoted the development of altitude sickness risk assessment from traditional empirical judgment to quantitative and model-based approaches, providing theoretical and technical support for the early identification of altitude sickness.

[0003] However, in actual clinical promotion and application among populations, the aforementioned existing technologies still have some technical shortcomings and cannot meet the needs for comprehensive, full-cycle, and refined risk assessment of altitude sickness. For example: Existing predictive models generally suffer from the problem of limited disease spectrum coverage. Most technical solutions only build models for a single type of high-altitude disease, such as acute altitude sickness or high-altitude pulmonary edema, and cannot simultaneously achieve integrated risk assessment for both mild acute altitude sickness and severe high-altitude pulmonary edema. This makes it difficult to adapt to the whole-cycle prevention and control scenario after people enter the high-altitude area, from the early stage of the disease to the progression of severe illness. Existing models have significant limitations in feature system construction. Some models rely solely on clinical symptoms and basic vital signs, lacking objective data support from laboratory test indicators. Other models focus only on a single category of physiological indicators, failing to systematically integrate a multivariate system of demographic characteristics, high-altitude exposure-related characteristics, clinical symptoms, vital signs, and multidimensional biochemical and blood routine test indicators. This makes it impossible to comprehensively characterize the multidimensional influencing factors of high-altitude disease, resulting in limited predictive accuracy and generalization ability among different populations. There is a significant imbalance between model performance and clinical applicability in existing technologies. On the one hand, although complex models based on ensemble learning show high numerical accuracy on specific datasets, their black-box nature leads to extremely poor interpretability, making it difficult to provide clinicians and examinees with a clear interpretation of risk factor weights and thus hindering their widespread application in primary healthcare settings and population self-assessment. On the other hand, existing simple models based on logistic regression do not construct a hierarchical, multi-dimensional risk assessment system through multivariate logistic regression. The variable selection process lacks systematicity and standardization, and does not fully consider the class imbalance caused by the low proportion of positive samples for altitude sickness in clinical data. This results in insufficient sensitivity in identifying severe altitude sickness and a high risk of missed diagnosis in high-risk populations. Most existing models lack differentiated adaptation designs for high-altitude exposure scenarios. They do not perform risk calibration for different altitude gradients, different modes of transportation to the plateau, and different exposure durations. They also fail to achieve a full-process assessment from predicting the probability of disease incidence to quantifying risk levels and matching individualized intervention strategies. Therefore, they cannot meet the needs of refined assessment of the risk of high-altitude diseases for different populations in different scenarios such as tourism, work, and mountaineering. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for assessing the risk of high-altitude diseases based on multivariate logistic regression, which aims to solve the problems of existing models having limited disease spectrum coverage, incomplete feature system, poor interpretability, and lack of scenario-based full-cycle risk stratification capabilities.

[0005] This invention is achieved through the following technical solution: A method for assessing the risk of altitude sickness based on multivariate logistic regression includes the following steps: Collect multidimensional characteristic data of the individuals to be evaluated; wherein, the multidimensional characteristic data includes demographic characteristic data, high-altitude exposure-related characteristic data, clinical symptom data, vital sign data, and laboratory test indicator data; the laboratory test indicator data includes blood routine test indicators and biochemical test indicators; Multidimensional feature data is input into a pre-defined multivariate logistic regression risk assessment model; the multivariate logistic regression risk assessment model includes a risk sub-model for mild acute altitude sickness and a risk sub-model for severe high-altitude pulmonary edema; both the risk sub-model for mild acute altitude sickness and the risk sub-model for severe high-altitude pulmonary edema are constructed based on feature variables corresponding to multidimensional feature data of the same dimension; during the training phase, the multivariate logistic regression risk assessment model uses synthetic minority class oversampling technology to handle class imbalance in positive samples of altitude sickness, and uses stepwise regression combined with clinical prior knowledge to screen the independent variables entering each sub-model; Using the risk sub-models for mild acute altitude sickness and severe high-altitude pulmonary edema, the first risk probability value for mild acute altitude sickness and the second risk probability value for severe high-altitude pulmonary edema of the individual to be evaluated under the current high-altitude exposure scenario are output respectively. Obtain the scene parameters corresponding to the current high-altitude exposure scenario, and perform scene-adaptive calibration on the first risk probability value and the second risk probability value based on the scene parameters to obtain the calibrated first risk probability value and the calibrated second risk probability value; wherein, the scene parameters include altitude gradient parameters, transportation mode parameters for entering the plateau, and plateau exposure duration parameters; The calibrated first risk probability value is compared with a preset set of mild illness risk stratification thresholds to obtain the mild illness risk level of the individual to be assessed. The calibrated second risk probability value is compared with a preset set of severe illness risk stratification thresholds to obtain the severe illness risk level of the individual to be assessed. An integrated risk assessment result is generated based on the mild illness risk level and the severe illness risk level. The integrated risk assessment result is used to characterize the risk level of the individual to be assessed in the current high-altitude exposure scenario, from the early stage of acute altitude sickness to the progression of severe illness.

[0006] Optionally, the specific process of collecting the multidimensional feature data of the individual to be evaluated is as follows: Obtain the identification information of the individual to be evaluated, and retrieve the pre-stored demographic data in the electronic health record based on the identification information; wherein, the demographic data includes age, gender, body mass index and history of cardiopulmonary disease; High-altitude exposure-related characteristic data and clinical symptom data of the individuals to be evaluated were collected using a standardized consultation form. The high-altitude exposure-related characteristic data included the planned altitude, mode of transportation to the high-altitude area, and the duration of current continuous exposure in the high-altitude environment. The clinical symptom data included headache severity rating, fatigue severity rating, dizziness frequency rating, and resting respiratory distress severity rating. Vital sign data of the individual to be evaluated are collected using vital sign monitoring equipment; wherein, the vital sign data includes resting heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate and fingertip oxygen saturation; According to the preset sample collection and testing procedures, venous blood samples are collected from the individuals to be evaluated, and routine blood tests and biochemical tests are performed to obtain the values ​​of the routine blood test indicators and the values ​​of the biochemical test indicators. The routine blood test indicators include red blood cell count, hemoglobin concentration, hematocrit, white blood cell count, and platelet count; the biochemical test indicators include serum sodium ion concentration, serum potassium ion concentration, serum creatinine concentration, and serum urea nitrogen concentration. Missing value identification and outlier removal were performed on the collected demographic characteristic data, plateau exposure-related characteristic data, clinical symptom data, vital sign data, blood routine test index data, and biochemical test index data. The processed data were then converted into numerical variables according to the preset variable coding rules, and all numerical variables were combined to form multidimensional feature data.

[0007] Optionally, the construction process of the multivariate logistic regression risk assessment model is as follows: A historical sample dataset of people exposed to high altitudes is obtained. Each sample in the historical sample dataset has a feature variable of the same dimension as the multidimensional feature data and a corresponding outcome label. The outcome labels include no altitude sickness, mild acute altitude sickness, and severe high-altitude pulmonary edema. Data preprocessing is performed on the historical sample dataset to identify and remove samples with outliers and missing values ​​exceeding a preset threshold for each feature variable, thus obtaining a preprocessed historical sample dataset. Synthetic minority class oversampling technique was used to oversample positive samples of mild acute altitude sickness and positive samples of severe high altitude pulmonary edema in the historical sample preprocessing dataset to generate a balanced training dataset. The number of samples without altitude sickness, samples of mild acute altitude sickness, and samples of severe high altitude pulmonary edema in the balanced training dataset was set according to a preset ratio. Based on the balanced training dataset, a two-way variable screening was performed on all feature variables using stepwise regression combined with clinical prior knowledge constraints, resulting in a first feature variable subset corresponding to mild acute altitude sickness and a second feature variable subset corresponding to severe high-altitude pulmonary edema. By fitting and training the multivariate logistic regression model with the first subset of feature variables, a first linear prediction equation with the logarithm of the probability of mild acute altitude sickness as the dependent variable is obtained, which serves as the risk sub-model for mild acute altitude sickness; by fitting and training the multivariate logistic regression model with the second subset of feature variables, a second linear prediction equation with the logarithm of the probability of severe high-altitude pulmonary edema as the dependent variable is obtained, which serves as the risk sub-model for severe high-altitude pulmonary edema. The regression coefficients, odds ratios, and confidence intervals of each characteristic variable in the mild acute altitude sickness risk sub-model and the severe high-altitude pulmonary edema risk sub-model are calculated respectively. Statistical significance and clinical rationality are verified for each characteristic variable. After the verification is passed, the mild acute altitude sickness risk sub-model and the severe high-altitude pulmonary edema risk sub-model are combined to form the multivariate logistic regression risk assessment model.

[0008] Optionally, the specific process of performing bidirectional variable screening on all feature variables based on the balanced training dataset using stepwise regression combined with clinical prior knowledge constraints is as follows: Use all feature variables in the balanced training dataset as the set of candidate independent variables; For each independent variable in the candidate set, univariate logistic regression analysis was performed with the outcome of mild acute altitude sickness and the outcome of severe high-altitude pulmonary edema as dependent variables. Independent variables with a significance level greater than the first preset threshold in the univariate analysis were removed to obtain a preliminary set of selected variables. In the initial set of selected variables, mandatory retention variables specified by prior clinical knowledge are marked. These mandatory retention variables include age, dyspnea severity grade at rest, and fingertip oxygen saturation. Using the initial set of selected variables as input, a bidirectional stepwise regression is performed. In each iteration, the statistical significance level of each variable in the current model and the change in the Akaike Information Criterion are calculated. The independent variables that meet the significance level and cause the Akaike Information Criterion to decrease the most are included according to the second preset threshold. The independent variables that do not meet the significance level and cause the Akaike Information Criterion to increase the most are removed according to the third preset threshold. After each step of the independent variable inclusion or removal operation, check whether the current model contains all the mandatory retention variables. If any mandatory retention variable is not included, force it to be added to the current model and continue iterating. When there are no independent variables that simultaneously meet the inclusion and exclusion criteria, the stepwise regression process is terminated, and the variables included in the current model constitute either the first feature variable subset or the second feature variable subset. Multicollinearity diagnosis is performed on the first and second subsets of feature variables respectively, and variables with variance inflation factors exceeding the fourth preset threshold are removed to obtain the final subset of feature variables used for sub-model fitting.

[0009] Optionally, the specific process of the scene adaptability calibration is as follows: The actual altitude of the individual to be evaluated is obtained, and the actual altitude is compared with a preset altitude gradient reference benchmark to calculate the altitude gradient calibration coefficient; wherein, the altitude gradient calibration coefficient is a linear mapping value of the ratio of the actual altitude to the reference benchmark altitude after logarithmic transformation. The transportation mode parameters of the individuals to be evaluated entering the plateau are obtained, and the transportation mode calibration factor is determined based on the differences in plateau exposure rates corresponding to different transportation modes; wherein, the transportation mode calibration factor is assigned a value according to the rate range of altitude rise during the process of entering the plateau. Obtain the duration of continuous exposure of the individual to be evaluated at the current altitude, and calculate the exposure duration calibration coefficient based on the physiological laws of high altitude acclimatization; wherein, the exposure duration calibration coefficient is a monotonically decreasing nonlinear function value with respect to the duration of continuous exposure. Based on the altitude gradient calibration coefficient, the mode of transportation calibration factor, and the exposure duration calibration coefficient, a scenario composite calibration factor is constructed. Input the first risk probability value into the preset first scenario calibration formula, and perform a weighted calculation on the first risk probability value and the scenario composite calibration factor to output the calibrated first risk probability value. The second risk probability value is input into the preset second scenario calibration formula, and the second risk probability value is weighted and calculated with the scenario composite calibration factor to output the calibrated second risk probability value. The expressions for the first scene calibration formula and the second scene calibration formula are shown below:

[0010]

[0011] in, and These are the first and second risk probability values ​​before calibration, respectively. and These are the first and second risk probability values ​​after calibration, respectively. The scenario sensitivity coefficient for mild acute altitude sickness; The scenario sensitivity coefficient for severe high-altitude pulmonary edema; and All were predetermined through regression analysis of different scenarios in historical sample data; For scene-based composite calibration factors; is the base of the natural logarithm; The scenario composite calibration factor The calculation expression is shown in the following formula:

[0012] in, This refers to the actual altitude. The preset altitude gradient reference height; The elevation gradient influence weighting coefficient; Calibration factor for transportation modes; Weighting coefficients for the influence of modes of transportation; This refers to the duration of continuous exposure at the current altitude. The constant representing the acclimatization rate at high altitudes; The weighting coefficient is affected by the exposure duration; , , and All of these are constants determined based on prior knowledge of plateau physiology and fitting with historical samples.

[0013] Optionally, the process for determining the preset set of risk stratification thresholds for mild cases and the preset set of risk stratification thresholds for severe cases is as follows: Obtain the predicted sequence of the first risk probability of mild acute altitude sickness and the predicted sequence of the second risk probability of severe high altitude pulmonary edema output by the multivariate logistic regression risk assessment model on the historical sample dataset, and obtain the true outcome label corresponding to each sample. Using the first and second risk probability prediction sequences as test variables, and the actual outcomes of mild acute altitude sickness and severe high-altitude pulmonary edema as state variables, receiver operating characteristic curves were plotted, and the Youden index corresponding to each probability cutoff point was calculated. The probability value corresponding to the maximum Yoden index on the operating characteristic curve of subjects with mild acute altitude sickness is determined as the threshold for classifying mild cases as medium-to-high risk. The lower bound of the probability interval where the sensitivity is greater than or equal to a preset first sensitivity threshold and the specificity is greater than or equal to a preset first specificity threshold is determined as the threshold for classifying mild cases as low-to-medium risk. The low-to-medium risk threshold and the medium-to-high risk threshold together constitute the risk stratification threshold set for mild cases. If the first risk probability value is less than the low-to-medium risk threshold, the risk level of the individual being assessed with mild acute altitude sickness is determined to be low. If the first risk probability value is greater than or equal to the low-to-medium risk threshold and less than the medium-to-high risk threshold, the risk level of the individual being assessed with mild acute altitude sickness is determined to be medium. If the first risk probability value is greater than or equal to the medium-to-high risk threshold, the risk level of the individual being assessed with mild acute altitude sickness is determined to be high. The minimum probability value on the subject operating characteristic curve of severe high-altitude pulmonary edema, under the constraint of sensitivity greater than or equal to a preset second sensitivity threshold, is determined as the high-risk screening threshold for severe pulmonary edema; the maximum probability value under the constraint of specificity greater than or equal to a preset second specificity threshold is determined as the low-risk exclusion threshold for severe pulmonary edema. The low-risk exclusion threshold and the high-risk screening threshold together constitute the risk stratification threshold set for severe pulmonary edema. If the second risk probability value is less than the low-risk exclusion threshold, the risk level of severe high-altitude pulmonary edema of the individual to be evaluated is determined to be low; if the second risk probability value is greater than or equal to the low-risk exclusion threshold and less than the high-risk screening threshold, the risk level of severe high-altitude pulmonary edema of the individual to be evaluated is determined to be intermediate; if the second risk probability value is greater than or equal to the high-risk screening threshold, the risk level of severe high-altitude pulmonary edema of the individual to be evaluated is determined to be high.

[0014] Optionally, after generating the integrated risk assessment results, the following steps are also included: Based on the mild and severe risk levels, an individualized intervention plan corresponding to the current risk level is retrieved from a pre-defined intervention strategy matching rule base. The intervention strategy matching rule base contains non-pharmacological interventions, drug prevention recommendations, and medical referral indications mapped to different combinations of mild and severe risk levels. The individualized intervention plan is combined with the integrated risk assessment results to generate a structured risk assessment report; the structured risk assessment report includes risk probability values, risk level identifiers, the contribution weights of each characteristic variable to the risk probability, and the individualized intervention plan; When newly generated vital sign data or clinical symptom data of an individual under assessment are detected during continuous high-altitude exposure, the newly generated data is used as incremental input to call the multivariate logistic regression risk assessment model to update the first risk probability value and the second risk probability value, and the integrated risk assessment result and the structured risk assessment report are iteratively refreshed based on the updated probability values.

[0015] Based on the same inventive concept, this invention also provides a high-altitude disease risk assessment system based on multivariate logistic regression, used to implement the aforementioned high-altitude disease risk assessment method based on multivariate logistic regression, comprising: The data acquisition module is used to collect multidimensional characteristic data of the individual to be evaluated. The multidimensional characteristic data includes demographic characteristic data, high altitude exposure-related characteristic data, clinical symptom data, vital sign data, and laboratory test indicator data. The laboratory test indicator data includes blood routine test indicators and biochemical test indicators. The model storage module is used to store pre-built multivariate logistic regression risk assessment models. The multivariate logistic regression risk assessment models include a risk sub-model for mild acute altitude sickness and a risk sub-model for severe high-altitude pulmonary edema. Both the risk sub-model for mild acute altitude sickness and the risk sub-model for severe high-altitude pulmonary edema are constructed based on feature variables corresponding to multidimensional feature data of the same dimension. During the training phase, synthetic minority class oversampling technology is used to process the class imbalance of positive samples of altitude sickness, and stepwise regression combined with clinical prior knowledge is used to screen the independent variables entering each sub-model. The risk assessment module, connected to the data acquisition module and the model storage module, is used to receive the multidimensional feature data, call the mild acute altitude sickness risk sub-model to output the first risk probability value of the individual to be assessed developing mild acute altitude sickness in the current high altitude exposure scenario, and call the severe high altitude pulmonary edema risk sub-model to output the second risk probability value of the individual to be assessed developing severe high altitude pulmonary edema. The scenario calibration module, connected to the risk assessment module, is used to obtain the scenario parameters corresponding to the current plateau exposure scenario, perform scenario adaptive calibration on the first risk probability value and the second risk probability value based on the scenario parameters, and output the calibrated first risk probability value and the calibrated second risk probability value. The scenario parameters include altitude gradient parameters, transportation mode parameters for entering the plateau, and plateau exposure duration parameters. The risk stratification assessment module, connected to the scenario calibration module, is used to compare the calibrated first risk probability value with a preset set of mild case risk stratification thresholds to obtain the mild case risk level of the individual to be assessed, compare the calibrated second risk probability value with a preset set of severe case risk stratification thresholds to obtain the severe case risk level of the individual to be assessed, and generate an integrated risk assessment result based on the mild case risk level and the severe case risk level. The intervention matching and report generation module is connected to the risk stratification assessment module. It is used to retrieve the corresponding individualized intervention plan from the preset intervention strategy matching rule base based on the mild disease risk level and the severe disease risk level, and merge the individualized intervention plan with the integrated risk assessment result to generate a structured risk assessment report. The dynamic update module is connected to the data acquisition module, the risk assessment module, the scenario calibration module, the risk stratification assessment module, and the intervention matching and report generation module, respectively. When new vital sign data or clinical symptom data of the individual to be assessed during continuous high-altitude exposure are detected, the newly generated data is used as incremental input to trigger the risk assessment module, the scenario calibration module, and the risk stratification assessment module to iteratively update the first risk probability value, the second risk probability value, and the integrated risk assessment result, and refresh the structured risk assessment report.

[0016] Based on the same inventive concept, the present invention also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the above-described method for assessing the risk of high altitude diseases based on multiple logistic regression.

[0017] Based on the same inventive concept, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method for assessing the risk of high-altitude diseases based on multivariate logistic regression.

[0018] The technical solution of the present invention has at least the following advantages and beneficial effects: By constructing risk sub-models for mild acute altitude sickness and severe high-altitude pulmonary edema, we can simultaneously output the risk probability of an individual developing mild and severe altitude sickness in the current high-altitude exposure scenario. This covers the entire prevention and control scenario from the early stage of acute altitude sickness to the progression of severe illness, making up for the technical deficiency of existing models that only target a single disease type, and providing full-chain risk identification capabilities for the clinical and public health fields.

[0019] By integrating demographic characteristics, high-altitude exposure-related characteristics, clinical symptoms, vital signs, and objective laboratory data including blood routine and biochemical test indicators, a multi-source heterogeneous variable system is constructed to comprehensively characterize the complex influencing factors of high-altitude disease. This avoids the prediction bias caused by existing models relying on single-type features or lacking objective test data, and enhances the prediction accuracy and generalization performance of the model under different population and environmental conditions.

[0020] By employing a multivariate logistic regression framework, this approach ensures that the risk probability output is intuitive and interpretable. It also effectively addresses the class imbalance problem caused by the low proportion of positive samples for altitude sickness by using synthetic minority oversampling technology. Combined with stepwise regression and clinical prior knowledge, it achieves scientific screening of feature variables. While maintaining the simplicity of the model, it significantly improves the sensitivity of identifying high-risk diseases such as severe high-altitude pulmonary edema, reduces the risk of missed diagnosis, and is more conducive to promotion and application in primary healthcare scenarios and individual self-assessment.

[0021] For different exposure scenarios such as different altitude gradients, modes of transportation to the plateau, and duration of exposure at the plateau, the initial risk probability is calibrated to adapt to the scenario, so that the risk assessment results can dynamically reflect the impact of actual environmental and behavioral factors on the risk of disease, and meet the needs of refined risk assessment in diverse application scenarios such as tourism, work, and mountaineering.

[0022] By pre-setting risk stratification thresholds for mild and severe cases, the calibrated probability values ​​are transformed into clear risk levels for mild and severe cases, and an integrated risk assessment result is generated. This achieves a closed-loop assessment process from predicting the probability of disease incidence to classifying risk levels and matching individualized intervention strategies, providing a standardized and operable decision-making basis for early screening, precise intervention, and rational allocation of medical resources for plateau diseases. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating the altitude sickness risk assessment method based on multivariate logistic regression, as described in an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of the high altitude disease risk assessment system based on multivariate logistic regression, according to an embodiment of the present invention. Detailed Implementation

[0024] The following is a detailed description of the embodiments, in conjunction with the accompanying drawings.

[0025] Reference Figure 1 A method for assessing the risk of altitude sickness based on multivariate logistic regression includes the following steps: Step 1: Collect multidimensional characteristic data of the individual to be evaluated; wherein, the multidimensional characteristic data includes demographic characteristic data, high altitude exposure-related characteristic data, clinical symptom data, vital sign data, and laboratory test indicator data; the laboratory test indicator data includes blood routine test indicators and biochemical test indicators.

[0026] In some embodiments, the specific process of collecting multidimensional feature data of the individual to be evaluated is as follows: Obtain the identification information of the individual to be evaluated, and retrieve the pre-stored demographic data in the electronic health record based on the identification information; wherein, the demographic data includes age, gender, body mass index and history of cardiopulmonary disease; High-altitude exposure-related characteristic data and clinical symptom data of the individuals to be evaluated were collected using a standardized consultation form. The high-altitude exposure-related characteristic data included the planned altitude, mode of transportation to the high-altitude area, and the duration of current continuous exposure in the high-altitude environment. The clinical symptom data included headache severity rating, fatigue severity rating, dizziness frequency rating, and resting respiratory distress severity rating. Vital sign data of the individual to be evaluated are collected using vital sign monitoring equipment; wherein, the vital sign data includes resting heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate and fingertip oxygen saturation; According to the preset sample collection and testing procedures, venous blood samples are collected from the individuals to be evaluated, and routine blood tests and biochemical tests are performed to obtain the values ​​of the routine blood test indicators and the values ​​of the biochemical test indicators. The routine blood test indicators include red blood cell count, hemoglobin concentration, hematocrit, white blood cell count, and platelet count; the biochemical test indicators include serum sodium ion concentration, serum potassium ion concentration, serum creatinine concentration, and serum urea nitrogen concentration. Missing value identification and outlier removal were performed on the collected demographic characteristic data, plateau exposure-related characteristic data, clinical symptom data, vital sign data, blood routine test index data, and biochemical test index data. The processed data were then converted into numerical variables according to the preset variable coding rules, and all numerical variables were combined to form multidimensional feature data.

[0027] Step 2: Input the multidimensional feature data into the preset multivariate logistic regression risk assessment model; the multivariate logistic regression risk assessment model includes a risk sub-model for mild acute altitude sickness and a risk sub-model for severe high-altitude pulmonary edema; both the risk sub-model for mild acute altitude sickness and the risk sub-model for severe high-altitude pulmonary edema are constructed based on feature variables corresponding to the multidimensional feature data of the same dimension; during the training phase, the multivariate logistic regression risk assessment model uses synthetic minority class oversampling technology to process the class imbalance of positive samples of altitude sickness, and uses stepwise regression combined with clinical prior knowledge to screen the independent variables entering each sub-model.

[0028] In some embodiments, the construction process of the multivariate logistic regression risk assessment model is as follows: A historical sample dataset of people exposed to high altitudes is obtained. Each sample in the historical sample dataset has a feature variable of the same dimension as the multidimensional feature data and a corresponding outcome label. The outcome labels include no altitude sickness, mild acute altitude sickness, and severe high-altitude pulmonary edema. Data preprocessing is performed on the historical sample dataset to identify and remove samples with outliers and missing values ​​exceeding a preset threshold for each feature variable, thus obtaining a preprocessed historical sample dataset. Synthetic minority class oversampling technique was used to oversample positive samples of mild acute altitude sickness and positive samples of severe high altitude pulmonary edema in the historical sample preprocessing dataset to generate a balanced training dataset. The number of samples without altitude sickness, samples of mild acute altitude sickness, and samples of severe high altitude pulmonary edema in the balanced training dataset was set according to a preset ratio. Based on the balanced training dataset, a two-way variable screening was performed on all feature variables using stepwise regression combined with clinical prior knowledge constraints, resulting in a first feature variable subset corresponding to mild acute altitude sickness and a second feature variable subset corresponding to severe high-altitude pulmonary edema. By fitting and training the multivariate logistic regression model using the first subset of feature variables, a first linear prediction equation is obtained, with the logarithm of the probability of mild acute altitude sickness as the dependent variable, serving as the risk sub-model for mild acute altitude sickness. Similarly, by fitting and training the multivariate logistic regression model using the second subset of feature variables, a second linear prediction equation is obtained, with the logarithm of the probability of severe high-altitude pulmonary edema as the dependent variable, serving as the risk sub-model for severe high-altitude pulmonary edema. The first linear prediction equation is shown below:

[0029] in, The log-odds is the probability of mild acute altitude sickness. This represents the probability of an individual experiencing mild acute altitude sickness under the current high-altitude exposure scenario, with a range of values. ; For the mild case sub-model, the intercept term represents the base level of the log odds when all independent variables are zero; The final number of independent variables included in the mild case sub-model; The first in the mild case sub-model The regression coefficients of the characteristic variables; The first selected for the mild case sub-model The actual values ​​of each characteristic variable (such as age, headache grade, fingertip blood oxygen saturation, etc.); The second linear prediction equation is shown below:

[0030] in, This represents the logarithmic probability of the occurrence of severe high-altitude pulmonary edema. This represents the probability of an individual developing severe high-altitude pulmonary edema under the current high-altitude exposure scenario, with a range of values. ; Intercept term of the severe illness sub-model; This represents the final number of independent variables included in the severe illness sub-model. The first in the severe illness sub-model The regression coefficients of the characteristic variables; The first selected for the critical care sub-model The actual values ​​of each characteristic variable; The regression coefficients, odds ratios, and confidence intervals of each characteristic variable in the mild acute altitude sickness risk sub-model and the severe high-altitude pulmonary edema risk sub-model are calculated respectively. Statistical significance and clinical rationality are verified for each characteristic variable. After the verification is passed, the mild acute altitude sickness risk sub-model and the severe high-altitude pulmonary edema risk sub-model are combined to form the multivariate logistic regression risk assessment model.

[0031] In some embodiments, the specific process of performing bidirectional variable screening on all feature variables based on the balanced training dataset using stepwise regression combined with clinical prior knowledge constraints is as follows: Use all feature variables in the balanced training dataset as the set of candidate independent variables; For each independent variable in the candidate set, univariate logistic regression analysis was performed with the outcome of mild acute altitude sickness and the outcome of severe high-altitude pulmonary edema as dependent variables. Independent variables with a significance level greater than the first preset threshold in the univariate analysis were removed to obtain a preliminary set of selected variables. In the initial set of selected variables, mandatory retention variables specified by prior clinical knowledge are marked. These mandatory retention variables include age, dyspnea severity grade at rest, and fingertip oxygen saturation. Using the initial set of selected variables as input, a bidirectional stepwise regression is performed. In each iteration, the statistical significance level of each variable in the current model and the change in the Akaike Information Criterion are calculated. The independent variables that meet the significance level and cause the Akaike Information Criterion to decrease the most are included according to the second preset threshold. The independent variables that do not meet the significance level and cause the Akaike Information Criterion to increase the most are removed according to the third preset threshold. After each step of the independent variable inclusion or removal operation, check whether the current model contains all the mandatory retention variables. If any mandatory retention variable is not included, force it to be added to the current model and continue iterating. When there are no independent variables that simultaneously meet the inclusion and exclusion criteria, the stepwise regression process is terminated, and the variables included in the current model constitute either the first feature variable subset or the second feature variable subset. Multicollinearity diagnosis is performed on the first and second subsets of feature variables respectively, and variables with variance inflation factors exceeding the fourth preset threshold are removed to obtain the final subset of feature variables used for sub-model fitting.

[0032] Step 3: Using the risk sub-models for mild acute altitude sickness and severe high-altitude pulmonary edema, output the first risk probability value for mild acute altitude sickness and the second risk probability value for severe high-altitude pulmonary edema for the individual to be evaluated under the current high-altitude exposure scenario.

[0033] In some embodiments, after the collection and preprocessing of multidimensional feature data are completed, the system simultaneously inputs the standardized numerical feature vectors into the trained sub-models for the risk of mild acute altitude sickness and severe high-altitude pulmonary edema. In the risk sub-model of mild acute altitude sickness, the model calculates the first risk probability value of mild acute altitude sickness for the individual under the current altitude exposure conditions by substituting the actual values ​​of characteristic variables such as age, gender, body mass index, history of cardiopulmonary disease, planned altitude, mode of transportation to the plateau, duration of continuous exposure, headache severity grade, fatigue severity grade, dizziness frequency grade, resting dyspnea severity grade, resting heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate, fingertip oxygen saturation, red blood cell count, hemoglobin concentration, hematocrit, white blood cell count, platelet count, serum sodium ion concentration, serum potassium ion concentration, serum creatinine concentration, and serum urea nitrogen concentration into the first linear prediction equation selected by stepwise regression combined with clinical prior knowledge constraints. The first linear prediction equation uses the logarithm of the probability of mild acute altitude sickness as the dependent variable. By performing an inverse exponential transformation on the linear combination of each characteristic variable and its corresponding regression coefficient, the model calculates the first risk probability value of mild acute altitude sickness for the individual under the current altitude exposure conditions. In the risk sub-model for severe high-altitude pulmonary edema, the model also reads the actual values ​​of all the aforementioned feature variables, but uses a subset of second feature variables independently selected for the severe high-altitude pulmonary edema outcome and the corresponding second linear prediction equation. Since the pathological mechanisms of severe high-altitude pulmonary edema differ from those of mild acute mountain sickness, the regression coefficients and significance levels of each variable in the second feature variable subset reflect the specific association strength of the severe outcome. The model also performs linear combination and probability transformation operations to output the second risk probability value for the individual being assessed to develop severe high-altitude pulmonary edema under the current high-altitude exposure conditions.

[0034] Both sub-models output continuous probability values ​​ranging from 0 to 1, quantifying the risk of mild and severe altitude sickness, respectively. This achieves integrated probability prediction for both mild acute altitude sickness and severe high-altitude pulmonary edema, providing a quantifiable data foundation for the full-cycle assessment from early warning of altitude sickness to prevention and control of severe progression. Because both sub-models employ synthetic minority class oversampling during training to address the issue of insufficient positive samples, the model's sensitivity in identifying rare outcomes such as severe high-altitude pulmonary edema is not reduced due to class imbalance, even in real-world clinical scenarios where the proportion of positive altitude sickness samples is low. This effectively avoids the risk of missed diagnoses of high-risk groups due to sample imbalance found in existing models.

[0035] Step 4: Obtain the scene parameters corresponding to the current high-altitude exposure scenario. Based on the scene parameters, perform scene-adaptive calibration on the first risk probability value and the second risk probability value respectively to obtain the calibrated first risk probability value and the calibrated second risk probability value. The scene parameters include altitude gradient parameters, transportation mode parameters for entering the plateau, and plateau exposure duration parameters.

[0036] In some embodiments, the specific process of scene adaptability calibration is as follows: The actual altitude of the individual to be evaluated is obtained, and the actual altitude is compared with a preset altitude gradient reference benchmark to calculate the altitude gradient calibration coefficient; wherein, the altitude gradient calibration coefficient is a linear mapping value of the ratio of the actual altitude to the reference benchmark altitude after logarithmic transformation. The transportation mode parameters of the individuals to be evaluated entering the plateau are obtained, and the transportation mode calibration factor is determined based on the differences in plateau exposure rates corresponding to different transportation modes; wherein, the transportation mode calibration factor is assigned a value according to the rate range of altitude rise during the process of entering the plateau. Obtain the duration of continuous exposure of the individual to be evaluated at the current altitude, and calculate the exposure duration calibration coefficient based on the physiological laws of high altitude acclimatization; wherein, the exposure duration calibration coefficient is a monotonically decreasing nonlinear function value with respect to the duration of continuous exposure. Based on the altitude gradient calibration coefficient, the mode of transportation calibration factor, and the exposure duration calibration coefficient, a scenario composite calibration factor is constructed. Input the first risk probability value into the preset first scenario calibration formula, and perform a weighted calculation on the first risk probability value and the scenario composite calibration factor to output the calibrated first risk probability value. The second risk probability value is input into the preset second scenario calibration formula, and the second risk probability value is weighted and calculated with the scenario composite calibration factor to output the calibrated second risk probability value. The expressions for the first scene calibration formula and the second scene calibration formula are shown below:

[0037]

[0038] in, and These are the first and second risk probability values ​​before calibration, respectively. and These are the first and second risk probability values ​​after calibration, respectively. The scenario sensitivity coefficient for mild acute altitude sickness; The scenario sensitivity coefficient for severe high-altitude pulmonary edema; and All were predetermined through regression analysis of different scenarios in historical sample data; For scene-based composite calibration factors; is the base of the natural logarithm; The scenario composite calibration factor The calculation expression is shown in the following formula:

[0039] in, This refers to the actual altitude. The preset altitude gradient reference height; The elevation gradient influence weighting coefficient; Calibration factor for transportation modes; Weighting coefficients for the influence of modes of transportation; This refers to the duration of continuous exposure at the current altitude. The constant representing the acclimatization rate at high altitudes; The weighting coefficient is affected by the exposure duration; , , and All of these are constants determined based on prior knowledge of plateau physiology and fitting with historical samples.

[0040] Step 5: Compare the calibrated first risk probability value with the preset mild illness risk stratification threshold set to obtain the mild illness risk level of the individual to be assessed. Compare the calibrated second risk probability value with the preset severe illness risk stratification threshold set to obtain the severe illness risk level of the individual to be assessed. Generate an integrated risk assessment result based on the mild illness risk level and the severe illness risk level. The integrated risk assessment result is used to characterize the risk level of the individual to be assessed in the current high-altitude exposure scenario, from the early stage of acute altitude sickness to the progression of severe illness.

[0041] In some embodiments, the process of determining the preset set of risk stratification thresholds for mild cases and the preset set of risk stratification thresholds for severe cases is as follows: Obtain the predicted sequence of the first risk probability of mild acute altitude sickness and the predicted sequence of the second risk probability of severe high altitude pulmonary edema output by the multivariate logistic regression risk assessment model on the historical sample dataset, and obtain the true outcome label corresponding to each sample. Using the first and second risk probability prediction sequences as test variables, and the actual outcomes of mild acute altitude sickness and severe high-altitude pulmonary edema as state variables, receiver operating characteristic curves were plotted, and the Youden index corresponding to each probability cutoff point was calculated. The probability value corresponding to the maximum Yoden index on the operating characteristic curve of subjects with mild acute altitude sickness is determined as the threshold for classifying mild cases as medium-to-high risk. The lower bound of the probability interval where the sensitivity is greater than or equal to a preset first sensitivity threshold and the specificity is greater than or equal to a preset first specificity threshold is determined as the threshold for classifying mild cases as low-to-medium risk. The low-to-medium risk threshold and the medium-to-high risk threshold together constitute the risk stratification threshold set for mild cases. If the first risk probability value is less than the low-to-medium risk threshold, the risk level of the individual being assessed with mild acute altitude sickness is determined to be low. If the first risk probability value is greater than or equal to the low-to-medium risk threshold and less than the medium-to-high risk threshold, the risk level of the individual being assessed with mild acute altitude sickness is determined to be medium. If the first risk probability value is greater than or equal to the medium-to-high risk threshold, the risk level of the individual being assessed with mild acute altitude sickness is determined to be high. The minimum probability value on the subject operating characteristic curve of severe high-altitude pulmonary edema, under the constraint of sensitivity greater than or equal to a preset second sensitivity threshold, is determined as the high-risk screening threshold for severe pulmonary edema; the maximum probability value under the constraint of specificity greater than or equal to a preset second specificity threshold is determined as the low-risk exclusion threshold for severe pulmonary edema. The low-risk exclusion threshold and the high-risk screening threshold together constitute the risk stratification threshold set for severe pulmonary edema. If the second risk probability value is less than the low-risk exclusion threshold, the risk level of severe high-altitude pulmonary edema of the individual to be evaluated is determined to be low; if the second risk probability value is greater than or equal to the low-risk exclusion threshold and less than the high-risk screening threshold, the risk level of severe high-altitude pulmonary edema of the individual to be evaluated is determined to be intermediate; if the second risk probability value is greater than or equal to the high-risk screening threshold, the risk level of severe high-altitude pulmonary edema of the individual to be evaluated is determined to be high.

[0042] In some embodiments, after generating the integrated risk assessment results, the following steps are also included: Based on the mild and severe risk levels, an individualized intervention plan corresponding to the current risk level is retrieved from a pre-defined intervention strategy matching rule base. The intervention strategy matching rule base contains non-pharmacological interventions, drug prevention recommendations, and medical referral indications mapped to different combinations of mild and severe risk levels. The individualized intervention plan is combined with the integrated risk assessment results to generate a structured risk assessment report; the structured risk assessment report includes risk probability values, risk level identifiers, the contribution weights of each characteristic variable to the risk probability, and the individualized intervention plan; When newly generated vital sign data or clinical symptom data of an individual under assessment are detected during continuous high-altitude exposure, the newly generated data is used as incremental input to call the multivariate logistic regression risk assessment model to update the first risk probability value and the second risk probability value, and the integrated risk assessment result and the structured risk assessment report are iteratively refreshed based on the updated probability values.

[0043] Based on the same inventive concept, and corresponding to any of the above embodiments, refer to... Figure 2 This invention provides a high-altitude disease risk assessment system based on multivariate logistic regression, used to implement the aforementioned high-altitude disease risk assessment method based on multivariate logistic regression, comprising: The data acquisition module is used to collect multidimensional characteristic data of the individual to be evaluated. The multidimensional characteristic data includes demographic characteristic data, high altitude exposure-related characteristic data, clinical symptom data, vital sign data, and laboratory test indicator data. The laboratory test indicator data includes blood routine test indicators and biochemical test indicators. The model storage module is used to store pre-built multivariate logistic regression risk assessment models. The multivariate logistic regression risk assessment models include a risk sub-model for mild acute altitude sickness and a risk sub-model for severe high-altitude pulmonary edema. Both the risk sub-model for mild acute altitude sickness and the risk sub-model for severe high-altitude pulmonary edema are constructed based on feature variables corresponding to multidimensional feature data of the same dimension. During the training phase, synthetic minority class oversampling technology is used to process the class imbalance of positive samples of altitude sickness, and stepwise regression combined with clinical prior knowledge is used to screen the independent variables entering each sub-model. The risk assessment module, connected to the data acquisition module and the model storage module, is used to receive the multidimensional feature data, call the mild acute altitude sickness risk sub-model to output the first risk probability value of the individual to be assessed developing mild acute altitude sickness in the current high altitude exposure scenario, and call the severe high altitude pulmonary edema risk sub-model to output the second risk probability value of the individual to be assessed developing severe high altitude pulmonary edema. The scenario calibration module, connected to the risk assessment module, is used to obtain the scenario parameters corresponding to the current plateau exposure scenario, perform scenario adaptive calibration on the first risk probability value and the second risk probability value based on the scenario parameters, and output the calibrated first risk probability value and the calibrated second risk probability value. The scenario parameters include altitude gradient parameters, transportation mode parameters for entering the plateau, and plateau exposure duration parameters. The risk stratification assessment module, connected to the scenario calibration module, is used to compare the calibrated first risk probability value with a preset set of mild case risk stratification thresholds to obtain the mild case risk level of the individual to be assessed, compare the calibrated second risk probability value with a preset set of severe case risk stratification thresholds to obtain the severe case risk level of the individual to be assessed, and generate an integrated risk assessment result based on the mild case risk level and the severe case risk level. The intervention matching and report generation module is connected to the risk stratification assessment module. It is used to retrieve the corresponding individualized intervention plan from the preset intervention strategy matching rule base based on the mild disease risk level and the severe disease risk level, and merge the individualized intervention plan with the integrated risk assessment result to generate a structured risk assessment report. The dynamic update module is connected to the data acquisition module, the risk assessment module, the scenario calibration module, the risk stratification assessment module, and the intervention matching and report generation module, respectively. When new vital sign data or clinical symptom data of the individual to be assessed during continuous high-altitude exposure are detected, the newly generated data is used as incremental input to trigger the risk assessment module, the scenario calibration module, and the risk stratification assessment module to iteratively update the first risk probability value, the second risk probability value, and the integrated risk assessment result, and refresh the structured risk assessment report.

[0044] Based on the same inventive concept, corresponding to any of the above embodiments, the present invention provides an electronic device, including a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the high-altitude disease risk assessment method based on multivariate logistic regression of the embodiments.

[0045] Alternatively, the aforementioned electronic device may be a server.

[0046] In addition, this embodiment also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the altitude disease risk assessment method based on multivariate logistic regression of the embodiment.

[0047] It is understood that the processor in the embodiments of the present invention may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0048] The method steps in the embodiments of the present invention can be implemented in hardware or by a processor executing software instructions. The software instructions can consist of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, portable hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can reside in an ASIC.

[0049] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a storage medium or transmitted through a storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive (SSD)).

Claims

1. A method for assessing the risk of altitude sickness based on multivariate logistic regression, characterized in that, Includes the following steps: Collect multidimensional characteristic data of the individuals to be evaluated; wherein, the multidimensional characteristic data includes demographic characteristic data, high-altitude exposure-related characteristic data, clinical symptom data, vital sign data, and laboratory test indicator data; the laboratory test indicator data includes blood routine test indicators and biochemical test indicators; Multidimensional feature data is input into a pre-defined multivariate logistic regression risk assessment model; the multivariate logistic regression risk assessment model includes a risk sub-model for mild acute altitude sickness and a risk sub-model for severe high-altitude pulmonary edema; both the risk sub-model for mild acute altitude sickness and the risk sub-model for severe high-altitude pulmonary edema are constructed based on feature variables corresponding to multidimensional feature data of the same dimension; during the training phase, the multivariate logistic regression risk assessment model uses synthetic minority class oversampling technology to handle class imbalance in positive samples of altitude sickness, and uses stepwise regression combined with clinical prior knowledge to screen the independent variables entering each sub-model; Using the risk sub-models for mild acute altitude sickness and severe high-altitude pulmonary edema, the first risk probability value for mild acute altitude sickness and the second risk probability value for severe high-altitude pulmonary edema of the individual to be evaluated under the current high-altitude exposure scenario are output respectively. Obtain the scene parameters corresponding to the current high-altitude exposure scenario, and perform scene-adaptive calibration on the first risk probability value and the second risk probability value based on the scene parameters to obtain the calibrated first risk probability value and the calibrated second risk probability value; wherein, the scene parameters include altitude gradient parameters, transportation mode parameters for entering the plateau, and plateau exposure duration parameters; The calibrated first risk probability value is compared with a preset set of mild illness risk stratification thresholds to obtain the mild illness risk level of the individual to be assessed. The calibrated second risk probability value is compared with a preset set of severe illness risk stratification thresholds to obtain the severe illness risk level of the individual to be assessed. An integrated risk assessment result is generated based on the mild illness risk level and the severe illness risk level. The integrated risk assessment result is used to characterize the risk level of the individual to be assessed in the current high-altitude exposure scenario, from the early stage of acute altitude sickness to the progression of severe illness.

2. The altitude sickness risk assessment method based on multivariate logistic regression as described in claim 1, characterized in that... The specific process for collecting multidimensional feature data of the individual to be evaluated is as follows: Obtain the identification information of the individual to be evaluated, and retrieve the pre-stored demographic data in the electronic health record based on the identification information; wherein, the demographic data includes age, gender, body mass index and history of cardiopulmonary disease; High-altitude exposure-related characteristic data and clinical symptom data of the individuals to be evaluated were collected using a standardized consultation form. The high-altitude exposure-related characteristic data included the planned altitude, mode of transportation to the high-altitude area, and the duration of current continuous exposure in the high-altitude environment. The clinical symptom data included headache severity rating, fatigue severity rating, dizziness frequency rating, and resting respiratory distress severity rating. Vital sign data of the individual to be evaluated are collected using vital sign monitoring equipment; wherein, the vital sign data includes resting heart rate, systolic blood pressure, diastolic blood pressure, respiratory rate and fingertip oxygen saturation; According to the preset sample collection and testing procedures, venous blood samples are collected from the individuals to be evaluated, and routine blood tests and biochemical tests are performed to obtain the values ​​of the routine blood test indicators and the values ​​of the biochemical test indicators. The routine blood test indicators include red blood cell count, hemoglobin concentration, hematocrit, white blood cell count, and platelet count; the biochemical test indicators include serum sodium ion concentration, serum potassium ion concentration, serum creatinine concentration, and serum urea nitrogen concentration. Missing value identification and outlier removal were performed on the collected demographic characteristic data, plateau exposure-related characteristic data, clinical symptom data, vital sign data, blood routine test index data, and biochemical test index data. The processed data were then converted into numerical variables according to the preset variable coding rules, and all numerical variables were combined to form multidimensional feature data.

3. The altitude sickness risk assessment method based on multivariate logistic regression as described in claim 1, characterized in that, The construction process of the multivariate logistic regression risk assessment model is as follows: A historical sample dataset of people exposed to high altitudes is obtained. Each sample in the historical sample dataset has a feature variable of the same dimension as the multidimensional feature data and a corresponding outcome label. The outcome labels include no altitude sickness, mild acute altitude sickness, and severe high-altitude pulmonary edema. Data preprocessing is performed on the historical sample dataset to identify and remove samples with outliers and missing values ​​exceeding a preset threshold for each feature variable, thus obtaining a preprocessed historical sample dataset. Synthetic minority class oversampling technique was used to oversample positive samples of mild acute altitude sickness and positive samples of severe high altitude pulmonary edema in the historical sample preprocessing dataset to generate a balanced training dataset. The number of samples without altitude sickness, samples of mild acute altitude sickness, and samples of severe high altitude pulmonary edema in the balanced training dataset was set according to a preset ratio. Based on the balanced training dataset, a two-way variable screening was performed on all feature variables using stepwise regression combined with clinical prior knowledge constraints, resulting in a first feature variable subset corresponding to mild acute altitude sickness and a second feature variable subset corresponding to severe high-altitude pulmonary edema. By fitting and training the multivariate logistic regression model with the first subset of feature variables, a first linear prediction equation with the logarithm of the probability of mild acute altitude sickness as the dependent variable is obtained, which serves as the risk sub-model for mild acute altitude sickness; by fitting and training the multivariate logistic regression model with the second subset of feature variables, a second linear prediction equation with the logarithm of the probability of severe high-altitude pulmonary edema as the dependent variable is obtained, which serves as the risk sub-model for severe high-altitude pulmonary edema. The regression coefficients, odds ratios, and confidence intervals of each characteristic variable in the mild acute altitude sickness risk sub-model and the severe high-altitude pulmonary edema risk sub-model are calculated respectively. Statistical significance and clinical rationality are verified for each characteristic variable. After the verification is passed, the mild acute altitude sickness risk sub-model and the severe high-altitude pulmonary edema risk sub-model are combined to form the multivariate logistic regression risk assessment model.

4. The altitude sickness risk assessment method based on multivariate logistic regression as described in claim 3, characterized in that, The specific process of performing bidirectional variable screening on all feature variables based on the balanced training dataset, using stepwise regression combined with clinical prior knowledge constraints, is as follows: Use all feature variables in the balanced training dataset as the set of candidate independent variables; For each independent variable in the candidate set, univariate logistic regression analysis was performed with the outcome of mild acute altitude sickness and the outcome of severe high-altitude pulmonary edema as dependent variables. Independent variables with a significance level greater than the first preset threshold in the univariate analysis were removed to obtain a preliminary set of selected variables. In the initial set of selected variables, mandatory retention variables specified by prior clinical knowledge were marked. These mandatory retention variables included age, dyspnea severity grade at rest, and fingertip oxygen saturation. Using the initial set of selected variables as input, a bidirectional stepwise regression is performed. In each iteration, the statistical significance level of each variable in the current model and the change in the Akaike Information Criterion are calculated. The independent variables that meet the significance level and cause the Akaike Information Criterion to decrease the most are included according to the second preset threshold. The independent variables that do not meet the significance level and cause the Akaike Information Criterion to increase the most are removed according to the third preset threshold. After each step of the independent variable inclusion or removal operation, check whether the current model contains all the mandatory retention variables. If any mandatory retention variable is not included, force it to be added to the current model and continue iterating. When there are no independent variables that simultaneously meet the inclusion and exclusion criteria, the stepwise regression process is terminated, and the variables included in the current model constitute either the first feature variable subset or the second feature variable subset. Multicollinearity diagnosis is performed on the first and second subsets of feature variables respectively, and variables with variance inflation factors exceeding the fourth preset threshold are removed to obtain the final subset of feature variables used for sub-model fitting.

5. The altitude sickness risk assessment method based on multivariate logistic regression as described in claim 1, characterized in that, The specific process of the scene adaptive calibration is as follows: The actual altitude of the individual to be evaluated is obtained, and the actual altitude is compared with a preset altitude gradient reference benchmark to calculate the altitude gradient calibration coefficient; wherein, the altitude gradient calibration coefficient is a linear mapping value of the ratio of the actual altitude to the reference benchmark altitude after logarithmic transformation. The transportation mode parameters of the individuals to be evaluated entering the plateau are obtained, and the transportation mode calibration factor is determined based on the differences in plateau exposure rates corresponding to different transportation modes; wherein, the transportation mode calibration factor is assigned a value according to the rate range of altitude rise during the process of entering the plateau. Obtain the duration of continuous exposure of the individual to be evaluated at the current altitude, and calculate the exposure duration calibration coefficient based on the physiological laws of high altitude acclimatization; wherein, the exposure duration calibration coefficient is a monotonically decreasing nonlinear function value with respect to the duration of continuous exposure. Based on the altitude gradient calibration coefficient, the mode of transportation calibration factor, and the exposure duration calibration coefficient, a scenario composite calibration factor is constructed. Input the first risk probability value into the preset first scenario calibration formula, and perform a weighted calculation on the first risk probability value and the scenario composite calibration factor to output the calibrated first risk probability value. The second risk probability value is input into the preset second scenario calibration formula, and the second risk probability value is weighted and calculated with the scenario composite calibration factor to output the calibrated second risk probability value. The expressions for the first scene calibration formula and the second scene calibration formula are shown below: in, and These are the first and second risk probability values ​​before calibration, respectively. and These are the first and second risk probability values ​​after calibration, respectively. The scenario sensitivity coefficient for mild acute altitude sickness; The scenario sensitivity coefficient for severe high-altitude pulmonary edema; and All were predetermined through regression analysis of different scenarios in historical sample data; For scene-based composite calibration factors; is the base of the natural logarithm; The scenario composite calibration factor The calculation expression is shown in the following formula: in, This refers to the actual altitude. This is the preset altitude gradient reference height; The elevation gradient influence weighting coefficient; Calibration factor for transportation modes; Weighting coefficients for the influence of modes of transportation; This refers to the duration of continuous exposure at the current altitude. The constant representing the acclimatization rate at high altitudes; The weighting coefficient is affected by the exposure duration; , , and All of these are constants determined based on prior knowledge of plateau physiology and fitting with historical samples.

6. The altitude sickness risk assessment method based on multivariate logistic regression as described in claim 1, characterized in that, The process for determining the preset risk stratification threshold set for mild cases and the preset risk stratification threshold for severe cases is as follows: Obtain the predicted sequence of the first risk probability of mild acute altitude sickness and the predicted sequence of the second risk probability of severe high altitude pulmonary edema output by the multivariate logistic regression risk assessment model on the historical sample dataset, and obtain the true outcome label corresponding to each sample. Using the first and second risk probability prediction sequences as test variables, and the actual outcomes of mild acute altitude sickness and severe high-altitude pulmonary edema as state variables, receiver operating characteristic curves were plotted, and the Youden index corresponding to each probability cutoff point was calculated. The probability value corresponding to the maximum Gedeng index on the operating characteristic curve of subjects with mild acute altitude sickness is determined as the threshold for classifying mild cases as medium-to-high risk. The lower bound of the probability interval where the sensitivity is greater than or equal to a preset first sensitivity threshold and the specificity is greater than or equal to a preset first specificity threshold is determined as the threshold for classifying mild cases as low-to-medium risk. The low-to-medium risk threshold and the medium-to-high risk threshold together constitute the set of risk stratification thresholds for mild cases. If the first risk probability value is less than the low-to-medium risk threshold, the risk level of mild acute altitude sickness of the individual to be evaluated is determined to be low. If the first risk probability value is greater than or equal to the low-to-medium risk threshold and less than the medium-to-high risk threshold, the risk level of mild acute altitude sickness of the individual to be evaluated is determined to be medium. If the first risk probability value is greater than or equal to the threshold for classifying mild cases as medium to high risk, then the risk level of the individual to be assessed for mild acute altitude sickness is determined to be high. The minimum probability value on the operating characteristic curve of subjects with severe high-altitude pulmonary edema, under the constraint of sensitivity greater than or equal to a preset second sensitivity threshold, is determined as the high-risk screening threshold for severe illness; the maximum probability value under the constraint of specificity greater than or equal to a preset second specificity threshold is determined as the low-risk exclusion threshold for severe illness. The low-risk exclusion threshold and the high-risk screening threshold together constitute the risk stratification threshold set for severe illness. If the second risk probability value is less than the low-risk exclusion threshold, the risk level of severe high-altitude pulmonary edema of the individual to be evaluated is determined to be low; if the second risk probability value is greater than or equal to the low-risk exclusion threshold and less than the high-risk screening threshold, the risk level of severe high-altitude pulmonary edema of the individual to be evaluated is determined to be intermediate. If the second risk probability value is greater than or equal to the high-risk screening threshold for severe illness, the risk level of the individual to be assessed for severe high-altitude pulmonary edema is determined to be high.

7. The altitude sickness risk assessment method based on multiple logistic regression as described in claim 1, characterized in that, After generating the integrated risk assessment results, the following steps are also included: Based on the mild and severe risk levels, an individualized intervention plan corresponding to the current risk level is retrieved from a pre-defined intervention strategy matching rule base. The intervention strategy matching rule base contains non-pharmacological interventions, drug prevention recommendations, and medical referral indications mapped to different combinations of mild and severe risk levels. The individualized intervention plan is combined with the integrated risk assessment results to generate a structured risk assessment report; the structured risk assessment report includes risk probability values, risk level identifiers, the contribution weights of each characteristic variable to the risk probability, and the individualized intervention plan; When newly generated vital sign data or clinical symptom data of an individual under assessment are detected during continuous high-altitude exposure, the newly generated data is used as incremental input to call the multivariate logistic regression risk assessment model to update the first risk probability value and the second risk probability value, and the integrated risk assessment result and the structured risk assessment report are iteratively refreshed based on the updated probability values.

8. A high-altitude disease risk assessment system based on multivariate logistic regression, used to implement the high-altitude disease risk assessment method based on multivariate logistic regression as described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to collect multidimensional characteristic data of the individual to be evaluated. The multidimensional characteristic data includes demographic characteristic data, high altitude exposure-related characteristic data, clinical symptom data, vital sign data, and laboratory test indicator data. The laboratory test indicator data includes blood routine test indicators and biochemical test indicators. The model storage module is used to store pre-built multivariate logistic regression risk assessment models. The multivariate logistic regression risk assessment models include a risk sub-model for mild acute altitude sickness and a risk sub-model for severe high-altitude pulmonary edema. Both the risk sub-model for mild acute altitude sickness and the risk sub-model for severe high-altitude pulmonary edema are constructed based on feature variables corresponding to multidimensional feature data of the same dimension. During the training phase, synthetic minority class oversampling technology is used to process the class imbalance of positive samples of altitude sickness, and stepwise regression combined with clinical prior knowledge is used to screen the independent variables entering each sub-model. The risk assessment module, connected to the data acquisition module and the model storage module, is used to receive the multidimensional feature data, call the mild acute altitude sickness risk sub-model to output the first risk probability value of the individual to be assessed developing mild acute altitude sickness in the current high altitude exposure scenario, and call the severe high altitude pulmonary edema risk sub-model to output the second risk probability value of the individual to be assessed developing severe high altitude pulmonary edema. The scenario calibration module, connected to the risk assessment module, is used to obtain the scenario parameters corresponding to the current plateau exposure scenario, perform scenario adaptive calibration on the first risk probability value and the second risk probability value based on the scenario parameters, and output the calibrated first risk probability value and the calibrated second risk probability value. The scenario parameters include altitude gradient parameters, transportation mode parameters for entering the plateau, and plateau exposure duration parameters. The risk stratification assessment module, connected to the scenario calibration module, is used to compare the calibrated first risk probability value with a preset set of mild case risk stratification thresholds to obtain the mild case risk level of the individual to be assessed, compare the calibrated second risk probability value with a preset set of severe case risk stratification thresholds to obtain the severe case risk level of the individual to be assessed, and generate an integrated risk assessment result based on the mild case risk level and the severe case risk level. The intervention matching and report generation module is connected to the risk stratification assessment module. It is used to retrieve the corresponding individualized intervention plan from the preset intervention strategy matching rule base based on the mild disease risk level and the severe disease risk level, and merge the individualized intervention plan with the integrated risk assessment result to generate a structured risk assessment report. The dynamic update module is connected to the data acquisition module, the risk assessment module, the scenario calibration module, the risk stratification assessment module, and the intervention matching and report generation module, respectively. When new vital sign data or clinical symptom data of the individual to be assessed during continuous high-altitude exposure are detected, the newly generated data is used as incremental input to trigger the risk assessment module, the scenario calibration module, and the risk stratification assessment module to iteratively update the first risk probability value, the second risk probability value, and the integrated risk assessment result, and refresh the structured risk assessment report.

9. An electronic device, characterized in that, The device includes a memory and a processor, the memory being used to store a computer program, and the processor running the computer program to cause the electronic device to perform the altitude sickness risk assessment method based on multiple logistic regression as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the altitude sickness risk assessment method based on multivariate logistic regression as described in any one of claims 1-7.

Images