Acute exacerbation dystrophy risk prediction model for patients with senile chronic obstructive pulmonary disease and construction method thereof

Abstract

The invention belongs to the field of nursing science, and particularly relates to an acute exacerbation malnutrition risk prediction model for an old chronic obstructive pulmonary disease patient and a construction method of the acute exacerbation malnutrition risk prediction model, which are used for establishing a rapid and accurate prediction model for acute exacerbation malnutrition risk screening and evaluation of the old chronic obstructive pulmonary disease patient. The invention discloses a model for predicting the risk of acute exacerbation dystrophy of an old chronic obstructive pulmonary disease patient. The risk prediction model is a multivariate Logistic regression model: P =-12.845 + 1.022 * age + 0.339 * ethnic + 0.077 * educated degree + 0.683 * gastrointestinal tract symptom + 0.035 * height-0.126 * body mass index + 1.045 * partial pressure of carbon dioxide-0.077 * chloride ion-0.017 * serum albumin-0.004 * prealbumin + 0.005 * hypersensitive C reaction protein + 0.479 * lymphocyte number. According to the acute exacerbation malnutrition risk prediction model for the patient with the chronic obstructive pulmonary disease of the aged and the construction method thereof, the prediction model for rapidly and accurately screening and evaluating the acute exacerbation malnutrition risk of the patient with the chronic obstructive pulmonary disease of the aged is established.

Description

Technical Field

[0001] The present invention relates to the field of nursing, and in particular to a malnutrition risk prediction model for elderly patients with chronic obstructive pulmonary disease during the acute exacerbation period and a construction method thereof. Background Art

[0002] The Global Initiative for Chronic Obstructive Pulmonary Disease (COPD) 2023 states that chronic obstructive pulmonary disease (COPD) is a heterogeneous lung condition characterized by chronic respiratory symptoms and persistent (often progressive) airflow obstruction due to airway and / or alveolar abnormalities. COPD is the third leading cause of death worldwide [2] and is a key disease to be prevented and treated in the Healthy China 2030 Action Plan. It seriously affects the quality of life of patients and imposes a heavy economic burden on patients, their families, and society.

[0003] The elderly are a high-risk group for COPD, with severe lung function impairment, low awareness of COPD, serious underdiagnosis, lack of specificity in clinical symptoms, and frequent comorbidities with multiple diseases. Furthermore, they have their own physiological and pathological characteristics, so the diagnosis and treatment of COPD in the elderly population faces challenges. Malnutrition is one of the major extrapulmonary complications in elderly patients with COPD and can independently affect the risk of future events. The overall incidence of COPD combined with malnutrition in the elderly is high, which can cause muscle wasting, worsening dyspnea, respiratory muscle dysfunction, and even respiratory failure. It is an important risk factor for acute exacerbation and hospitalization of elderly patients with COPD. The more severe the malnutrition, the more likely it is to develop cachexia, which affects the quality of life, the number of acute attacks, and the number of hospitalizations, leading to an increase in mortality.

[0004] The American Thoracic Society states that COPD is primarily divided into acute exacerbations and stable phases. Studies indicate that the global inpatient mortality rate for patients with acute exacerbations of COPD is approximately 7%. Therefore, early screening, assessment, and standardized diagnosis and treatment of malnutrition in elderly COPD patients during acute exacerbations are particularly important. International and domestic guidelines for screening and assessment of malnutrition in elderly COPD patients recommend nutritional risk screening for all critically ill respiratory patients. The primary screening tools recommended in China are the Nutritional Risk in Critical Illness (NUTRIC) score and the Nutritional Risk Screening 2002 (NRS-2002) score. Conventional assessment tools are mostly designed for the general population, lack specificity, and are subject to subjective assessment by the assessor. Research has demonstrated that preventive management based on objective risk prediction has great potential to reduce the future risk and burden of COPD. In recent years, the majority of mainstream disease risk prediction tools developed internationally have been based on risk prediction models. Disease risk prediction models are statistical assessments based on disease risk factors, categorized by impact, and mathematically calculated using mathematical formulas to calculate the probability of a specific event occurring in a specific population in the future. A well-developed prediction model can accurately predict the risk of disease onset, helping clinicians identify high-risk patients and enabling close monitoring and effective management to reduce disease incidence. Rapid and accurate prediction models are urgently needed to screen and assess nutritional risk in elderly COPD patients during acute exacerbations, addressing the shortcomings of existing research. Summary of the Invention

[0005] The purpose of the present invention is to provide a malnutrition risk prediction model for elderly patients with chronic obstructive pulmonary disease during acute exacerbation and a construction method thereof, and to establish a fast and accurate prediction model for nutritional risk screening and assessment in elderly patients with chronic obstructive pulmonary disease during acute exacerbation.

[0006] In order to solve the above technical problems, the present invention provides a malnutrition risk prediction model for elderly patients with chronic obstructive pulmonary disease in the acute exacerbation period and a construction method thereof as follows:

[0007] A risk prediction model for malnutrition in elderly patients with chronic obstructive pulmonary disease during acute exacerbation is disclosed. The risk prediction model is a multivariate logistic regression model: P = -12.845 + 1.022 × age + 0.339 × ethnicity + 0.077 × education level + 0.683 × gastrointestinal symptoms + 0.035 × height - 0.126 × body mass index + 1.045 × carbon dioxide partial pressure - 0.077 × chloride ion - 0.017 × serum albumin - 0.004 × prealbumin + 0.005 × high-sensitivity C-reactive protein + 0.479 × lymphocyte count.

[0008] Optionally, the risk prediction model is expressed using a nomogram, including: score, age, ethnicity, education level, presence or absence of gastrointestinal symptoms, height, body mass index, carbon dioxide partial pressure, chloride ion, serum albumin, prealbumin, high-sensitivity C-reactive protein, lymphocyte count, total score, linear prediction value and risk probability of the disease.

[0009] Optionally, the risk prediction model is to express the multivariate logistic regression model in an online web calculator; or the risk prediction nomogram is to express it in an online web calculator.

[0010] The present invention also provides a method for constructing a risk prediction model for malnutrition in elderly patients with chronic obstructive pulmonary disease during acute exacerbations. The method comprises selecting, from clinical baseline data of a normal nutrition group and a malnutrition group of elderly patients with chronic obstructive pulmonary disease during acute exacerbations, indicators with statistically significant differences between the normal nutrition group and the malnutrition group as risk prediction variables for malnutrition in elderly patients with chronic obstructive pulmonary disease during acute exacerbations.

[0011] The risk prediction variables were subjected to Lasso regression analysis, and 12 variables were screened out from the risk prediction variables through the outcome effect of the Lasso regression path and Lasso cross-validation as risk prediction factors of the risk prediction model;

[0012] Using the risk prediction factors to construct a multivariate logistic regression model, i.e., a risk prediction model;

[0013] The multivariate logistic regression model was: P = -12.845 + 1.022 × age + 0.339 × ethnicity + 0.077 × education level + 0.683 × gastrointestinal symptoms + 0.035 × height - 0.126 × body mass index + 1.045 × partial pressure of carbon dioxide - 0.077 × chloride ion - 0.017 × serum albumin - 0.004 × prealbumin + 0.005 × high-sensitivity C-reactive protein + 0.479 × lymphocyte count.

[0014] Optionally, the risk prediction model is expressed by a nomogram, which includes: score, age, ethnicity, education level, presence or absence of gastrointestinal symptoms, height, body mass index, carbon dioxide partial pressure, chloride ion, serum albumin, prealbumin, high-sensitivity C-reactive protein, lymphocyte count, total score, linear prediction value and risk probability of the disease.

[0015] Optionally, after the risk prediction model is established, the multivariate logistic regression model is made into an online web calculator expression, or the risk prediction nomogram is made into an online web calculator expression.

[0016] Optionally, the elderly patients with acute exacerbation of COPD who meet the research criteria are: ① aged ≥ 60 years; ② meet the diagnostic criteria for COPD in the 2023 edition of the "Global Initiative for Chronic Obstructive Pulmonary Disease: Global Strategy for the Diagnosis, Treatment and Prevention of COPD"; ③ symptoms such as cough, sputum, chest tightness, and asthma have worsened, or are accompanied by acute infection symptoms such as fever, and require a change in medication regimen; ④ are conscious and able to communicate normally; ⑤ give informed consent and agree to cooperate with the investigator.

[0017] Optionally, the exclusion study criteria for elderly patients with acute exacerbation of COPD are: ① those diagnosed with other lung diseases such as lung cancer, bronchial asthma, interstitial lung disease, etc.; ② patients with combined heart failure or heart failure; ③ patients with combined metabolic diseases such as hyperthyroidism, hypothyroidism, diabetes, etc.; ④ patients with digestive system diseases within 3 months before hospitalization, or those with nutritional intervention treatment; ⑤ patients with a history of mental and psychological diseases.

[0018] Optionally, among the indicators selected that have statistical differences between the normal nutrition group and the malnutrition group, the statistical difference refers to P≤0.001.

[0019] Optionally, after the risk prediction model is established, it is verified by a machine learning method, and the machine learning method is linear regression, gradient boosting machine, support vector machine and random forest algorithm.

[0020] The malnutrition risk prediction model and construction method for elderly patients with chronic obstructive pulmonary disease during acute exacerbation provided by the present invention are intended to screen the risk factors of malnutrition in elderly patients with COPD during acute exacerbation, construct a malnutrition risk prediction model for elderly patients with COPD during acute exacerbation, promote the advancement of malnutrition risk screening for elderly patients with COPD during acute exacerbation, explore its application value in malnutrition risk prediction for elderly patients with COPD, and further transform it into a scientific and accurate malnutrition screening tool for elderly patients with COPD in combination with clinical characteristics and needs, such as a web calculator, to enable medical staff to carry out nutritional screening and health education for elderly patients with COPD during acute exacerbation at an early stage, take intervention measures as early as possible, and provide a theoretical basis for reducing the incidence of malnutrition risk and COPD mortality in such patients.

[0021] The present invention first screens for statistically significant variables in the nutrition group and the malnutrition group, then performs LASSO regression analysis. Using Lasso coefficient paths and Lasso cross-validation, 12 variables are selected. Multivariate logistic regression analysis is then performed on these 12 variables to generate a multivariate logistic regression equation containing all 12 variables. This method uses only LASSO regression analysis to select variables, resulting in a higher degree of discrimination than the multivariate logistic regression analysis used in the prior art.

[0022] The established malnutrition risk prediction model for elderly patients with acute exacerbations of chronic obstructive pulmonary disease (COPD) has demonstrated excellent sensitivity, discrimination, calibration, and clinical applicability. Using four algorithms, linear regression, gradient boosting, support vector machine, and random forest, the model demonstrated high precision, accuracy, and recall, with the gradient boosting algorithm demonstrating particularly high performance on validation data. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is the curve of Lasso regression coefficient changing with log(λ);

[0024] Figure 2 This is a graph showing the cross-validation model fit changing with log(λ);

[0025] Figure 3 It is a nomogram expression diagram of the risk prediction model of the present invention;

[0026] Figure 4 It is a web calculator expression diagram of the risk prediction model of the present invention;

[0027] Figure 5 is a ROC curve diagram of the risk prediction model of the present invention;

[0028] Figure 6 is a calibration graph of the risk prediction model of the present invention;

[0029] Figure 7 It is a decision curve analysis diagram of the risk prediction model of the present invention;

[0030] Figure 8 is the ROC curve diagram of the validation set risk prediction model of the present invention;

[0031] Figure 9 is a calibration graph of the validation set risk prediction model of the present invention;

[0032] Figure 10 It is a decision curve diagram of the validation set risk prediction model of the present invention;

[0033] Figure 11 This is a general linear regression algorithm model diagram in the fourth embodiment of the present invention;

[0034] Figure 12 This is a confusion matrix model diagram for internal verification of the general linear regression algorithm in the fourth embodiment of the present invention;

[0035] Figure 13 This is a model diagram of the gradient boosting machine algorithm in the fourth embodiment of the present invention;

[0036] Figure 14 This is a confusion matrix model diagram for internal verification of the GBM algorithm in the fourth embodiment of the present invention;

[0037] Figure 15 This is a confusion matrix model diagram for internal verification of the support vector machine algorithm in the fourth embodiment of the present invention;

[0038] Figure 16 It is the confusion matrix model for internal verification of the support vector machine algorithm in the fourth embodiment of the present invention;

[0039] Figure 17 It is the confusion matrix model for internal verification of the support vector machine algorithm in the fourth embodiment of the present invention;

[0040] Figure 18 It is the confusion matrix model for internal verification of the support vector machine algorithm in the fourth embodiment of the present invention;

[0041] Figure 19 This is a model diagram of the random forest algorithm in the fourth embodiment of the present invention;

[0042] Figure 20 This is the confusion matrix model for internal verification of the random forest algorithm in the fourth embodiment of the present invention;

[0043] Figure 21 is a ROC curve diagram of the Full-Logistic model in the comparative example of the present invention;

[0044] Figure 22 It is the ROC curve of the Lasso-Logistic model in the comparative example of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail in the following examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0046] The present invention provides a method for constructing a malnutrition risk prediction model for elderly patients with chronic obstructive pulmonary disease in the acute exacerbation period. The method first screens statistical indicators for a normal nutrition group and a malnutrition group, then performs Lasso regression analysis on the indicators to screen out 12 indicators as risk prediction factors, and constructs a multivariate logistic regression model. The multivariate logistic regression model is a numerical model. To facilitate observation, the present invention constructs a nomogram to express the risk prediction model based on the numerical model. To facilitate use by ordinary patients, the present invention also constructs a web calculator.

[0047] The present invention provides a method for constructing a malnutrition risk prediction model for elderly patients with chronic obstructive pulmonary disease during acute exacerbation, as follows:

[0048] From the clinical baseline data of the normal nutrition group and the malnutrition group in the elderly patients with COPD during acute exacerbation, the indicators with statistical differences between the normal nutrition group and the malnutrition group were selected as risk predictors of malnutrition in the elderly patients with COPD during acute exacerbation.

[0049] The risk prediction variables are subjected to Lasso regression analysis, and 12 variables are screened out from the risk prediction variables through Lasso coefficient path and Lasso cross validation as risk prediction factors of the risk prediction model;

[0050] Using the risk prediction factors to construct a multivariate logistic regression model, i.e., a risk prediction model;

[0051] The constructed multivariate logistic regression model was: P = -12.845 + 1.022 × age + 0.339 × ethnicity + 0.077 × education level + 0.683 × gastrointestinal symptoms + 0.035 × height - 0.126 × body mass index + 1.045 × partial pressure of carbon dioxide - 0.077 × chloride ion - 0.017 × serum albumin - 0.004 × prealbumin + 0.005 × high-sensitivity C-reactive protein + 0.479 × lymphocyte count.

[0052] The constructed risk prediction model was expressed through a nomogram, which included score, age, ethnicity, education level, presence or absence of gastrointestinal symptoms, height, body mass index, partial pressure of carbon dioxide, chloride ion, serum albumin, prealbumin, high-sensitivity C-reactive protein, lymphocyte count, total score, linear prediction value, and risk probability of disease.

[0053] In order to facilitate the use and operation of the model, after the risk prediction model is established, the multivariate logistic regression model is made into an online web calculator expression, or the risk prediction nomogram is made into an online web calculator expression.

[0054] Example 1 Training set risk prediction model construction

[0055] Before starting the present invention, the inventors of the present invention investigated elderly patients with acute exacerbation of chronic obstructive pulmonary disease who were admitted to the Respiratory Ward 1 and Ward 2 of the Respiratory and Critical Care Center of the First Affiliated Hospital of Henan University of Science and Technology from March to May 2022, eliminated invalid cases, and calculated that the incidence of malnutrition in elderly patients with acute exacerbation of COPD was 17.24%, 15.21% and 12.76%, respectively.

[0056] 101. From the clinical baseline data of the normal nutrition group and the malnutrition group in elderly patients with COPD during acute exacerbations, select indicators with statistically significant differences between the normal nutrition group and the malnutrition group as risk predictors of malnutrition in elderly patients with COPD during acute exacerbations; the specific indicators are as follows:

[0057] A total of 1571 elderly patients with chronic obstructive pulmonary disease who were admitted to the First Affiliated Hospital of Henan University of Science and Technology from January 2020 to December 2022 and met the criteria were selected as research subjects. According to the 70% and 30% allocation principle of training set and validation set, they were randomly divided into training set (n=966) and validation set (n=414) using SPSS software. The model was established with the training set and verified with the validation set.

[0058] Inclusion criteria for the research subjects: ① Age ≥ 60 years old; ② Meet the diagnostic criteria for COPD in the 2023 edition of the "Global Initiative for Chronic Obstructive Pulmonary Disease: Global Strategy for the Diagnosis, Treatment and Prevention of COPD"; ③ Those with worsening symptoms such as cough, sputum, chest tightness, and wheezing, or accompanied by acute infection symptoms such as fever, who need to change the medication regimen; ④ Those who are conscious and able to communicate normally; ⑤ Those who have informed consent and agree to cooperate with the investigation.

[0059] Exclusion criteria for the study subjects: ① Patients diagnosed with other lung diseases such as lung cancer, bronchial asthma, interstitial lung disease, etc.; ② Patients with combined heart failure or heart insufficiency; ③ Patients with combined metabolic diseases such as hyperthyroidism, hypothyroidism, diabetes, etc.; ④ Patients with digestive system diseases or nutritional intervention treatment within 3 months before hospitalization; ⑤ Patients with a history of mental and psychological diseases.

[0060] A total of 1,571 patients' clinical data were included in this study, covering 952 elderly patients with COPD who met the inclusion criteria and were admitted to a tertiary hospital in Henan Province from January 2021 to December 2022.

[0061] Among the patients participating in the study, 1,263 were male, accounting for 80.39%; 308 were female, accounting for 19.61%. According to the NRS 2002 score, 1,260 patients, accounting for 80.20%, had a score less than 3, indicating normal nutrition. The average age was 73.80 ± 7.73 years, the average weight was 60.72 ± 10.86 kg, and the average BMI was 21.46 ± 8.78 kg / m 2. There were 311 cases with a score greater than or equal to 3 points, that is, malnutrition, accounting for 19.80%. The average age was 77.29±9.10 years, the average weight was 57.34±11.46kg, and the average BMI was 19.40±3.51kg / m2. During the acute exacerbation period of elderly patients with COPD, the incidence of malnutrition was 19.80%. There were 867 patients with gastrointestinal symptoms, accounting for 55.19%; there were 704 patients without gastrointestinal symptoms, accounting for 44.81%. In terms of educational level, the number of people with primary school education was the largest, totaling 579 cases, accounting for 36.86%. Specific patient information is shown in Table 1, and patient clinical data is shown in Table 2.

[0062] Table 1 Clinical data of elderly patients with COPD Univariate analyses

[0063]

[0064]

[0065] Table 2 Univariate analyses of clinical data of elderly patients with COPD

[0066]

[0067]

[0068] Note: *Fisher's exact test

[0069] For the acute exacerbation of elderly COPD patients, the Fisher exact test of the present invention analyzed the clinical baseline data between the normal nutrition group and the malnutrition group in the training set, including age, gender, ethnicity, education level, gastrointestinal symptoms, height, weight, body mass index, oxygen partial pressure, carbon dioxide partial pressure, sodium ion, potassium ion, chloride ion, total protein, serum albumin, prealbumin, high-sensitivity C-reactive protein, hemoglobin, and lymphocyte count. Univariate analysis showed that the following indicators were statistically significant between the normal nutrition group and the malnutrition group: age (P < 0.001), education level (P = 0.001), presence or absence of gastrointestinal symptoms (P < 0.001), height (P < 0.001), weight (P < 0.001), body mass index (P < 0.001), partial pressure of carbon dioxide (P < 0.001), sodium ion (P < 0.001*), chloride ion (P < 0.001), serum albumin (P < 0.001), prealbumin (P < 0.001), and high-sensitivity C-reactive protein (P < 0.001). No other indicators were statistically significant between the two groups (P > 0.05).

[0070] 102. Perform Lasso regression analysis on the risk prediction variables, and select 12 variables from the risk prediction variables through Lasso coefficient path and Lasso cross-validation to serve as risk prediction factors of the risk prediction model;

[0071] The 101 statistically significant variables were processed and Lasso regression analysis was used to screen the risk predictors of malnutrition in elderly patients with COPD during acute exacerbation. Figure 1 It is a curve graph of the Lasso regression coefficient changing with log(λ), which represents the outcome effect coefficient corresponding to different variables in each Lasso path diagram. The smaller λ is, the more the curve of each variable tends to the horizontal axis, and the fewer independent variables have non-zero effect coefficients. Figure 2 The dashed line on the left represents the penalty coefficient X-min for minimum squared error, and the dashed line on the right represents the penalty coefficient Mse for a standard deviation from X-min, indicating the minimum number of variables that can be included in the model while maintaining statistical significance. Lasso cross-validation (K=10) shows that the number of model variables is 12 when log(λ) takes the minimum Mse cutoff, and 7 when it takes the minimum distance from the minimum value by 1 times the SE cutoff.

[0072] 103. The 12 risk predictors screened by Lasso regression were included in the multivariate logistic regression analysis;

[0073] The results, as shown in Tables 3 and 4, show that six variables—age (P < 0.001), presence of gastrointestinal symptoms (P = 0.014), body mass index (P = 0.001), partial pressure of carbon dioxide (P < 0.001), prealbumin (P = 0.026), and lymphocyte count (P = 0.009)—were independent risk factors for acute exacerbations in elderly patients with chronic obstructive pulmonary disease. P < 0.05 was considered statistically significant. Statistical analysis revealed no significant linear correlations among the remaining six variables.

[0074] Table 3 Multivariate logistic regression analyses

[0075]

[0076] Table 4 Multivariate logistic regression analyses

[0077]

[0078] Note: B: partial regression coefficient; SE: standard error of partial regression coefficient; Wald: Wald statistic; OR: odds ratio.

[0079] 104. Establishing a risk prediction model

[0080] A multivariate regression equation was established based on the above risk prediction factors, where P is the probability of malnutrition in elderly COPD patients during acute exacerbation. A multivariate logistic regression model was constructed:

[0081] P = -12.845 + 1.022 × age + 0.339 × ethnicity + 0.077 × education level + 0.683 × gastrointestinal symptoms + 0.035 × height - 0.126 × body mass index + 1.045 × partial pressure of carbon dioxide - 0.077 × chloride ion - 0.017 × serum albumin - 0.004 × prealbumin + 0.005 × high-sensitivity C-reactive protein + 0.479 × lymphocyte count

[0082] The risk factors and variable assignments for malnutrition in elderly COPD patients during acute exacerbation are shown in Table 5 , and the remaining variables were substituted into the equation according to the actual test values.

[0083] Table 5 Risk factors and variable assignment of malnutrition during acute exacerbation in elderly patients with COPD

[0084] COPD patients

[0085]

[0086] 105. Risk prediction model expressed through nomogram

[0087] The model in 104 is based on mathematical formulas. The calculation and operation of malnutrition in elderly COPD patients during acute exacerbation is relatively complex. The nomogram can make the complex logistic regression model more intuitive, more convenient, easier to understand, and more practical. Therefore, the nomogram model was constructed using R software. Figure 3 This nomogram includes 12 risk prediction factors, namely: age, ethnicity, education level, presence or absence of gastrointestinal symptoms, height, body mass index, partial pressure of carbon dioxide, chloride ion, serum albumin, prealbumin, high-sensitivity C-reactive protein, and lymphocyte count.

[0088] Assume a male patient, 65 years old, Han nationality, elementary school graduate, lack of appetite, height 175cm, body mass index 18kg / m 2 Laboratory examination showed: carbon dioxide partial pressure 65 mmHg, chloride ion 80 mmol / L, serum albumin 40 g / L, prealbumin 90 g / L, high-sensitivity C-reactive protein 200 mg / L, lymphocyte count 5.5×10 9 / L, the attending physician urgently needs to assess whether this male patient is at risk of malnutrition. After further questioning and obtaining medical records, it was found that the patient had no history of hypertension or diabetes, but had a long history of smoking. Figure 3 Each score is 11 points, 0 points, 2.5 points, 17.5 points, 25 points, 50 points, 25 points, 2.5 points, 5 points, 65 points, 22.5 points, 65 points, and the total score is 291 points. The corresponding malnutrition prevalence rate in elderly patients with COPD during acute exacerbation is >90%. After evaluation, the patient has a high risk of malnutrition. For specific examples, see Figure 3 In summary, the non-parametric risk prediction model is more intuitive, faster and more convenient to operate in disease diagnosis.

[0089] 106. Risk prediction model expressed through web calculator

[0090] Based on the mathematical formula in 104, a web calculator is generated. The address is http: / / www.redsuntech.cn:13000 / . Click the link to see the following Figure 4 The interface is displayed. Enter the actual test value, age, ethnicity, and education level. Fill in the actual values. The web calculator will assign values ​​to variables without specific values ​​according to the assignment in Table 5. For example, the patient input in the present invention is: male patient, 65 years old, Han nationality, elementary school graduate, lack of appetite, height 175 cm, body mass index 18 kg / m 2 Laboratory examination showed: carbon dioxide partial pressure 65 mmHg, chloride ion 80 mmol / L, serum albumin 40 g / L, prealbumin 90 g / L, high-sensitivity C-reactive protein 200 mg / L, lymphocyte count 5.5×10 9 / L, enter these variables in the web calculator, click Calculate Risk Value, and the current calculation result will be displayed, showing whether it is low, medium, or high risk, with corresponding prompts. This is convenient for ordinary patients to operate.

[0091] Example 2 Evaluation of the training set risk prediction model

[0092] The present invention also evaluates the established risk prediction model, as follows:

[0093] Model evaluation can determine the predictive effect of the model and reflect its clinical application value. Common performance evaluation indicators of the model include sensitivity, specificity, Youden index, and AUC value.

[0094] Sensitivity, specificity, Youden index and AUC value

[0095] Sensitivity: mainly refers to the proportion of patients who can correctly screen out actual positive results and identify them as positive; Specificity: mainly refers to the proportion of patients who can correctly screen out actual negative results and identify them as negative; Youden Index is a method to evaluate the authenticity of the screening test. Assuming that the harmfulness of false negatives (missed diagnosis rate) and false positives (misdiagnosis rate) is of equal significance, the Youden Index can be applied. The Youden Index is the sum of the sensitivity and specificity minus 1, which represents the total ability of the screening method to detect real patients and non-patients. The index value range is 0 to 1. The larger the index, the better the effect of the screening test and the greater the authenticity. The risk prediction model constructed in this study has a sensitivity of 0.764, a specificity of 0.744, and a Youden Index of 0.508. The risk prediction model of the present invention has good sensitivity, specificity, Youden Index and AUC values.

[0096] (2) Discrimination: The area under the ROC curve (AUC) can be used to evaluate the discriminative ability of the risk prediction model. The risk prediction model of the present invention plots the ROC curve, see Figure 5 The area under the ROC curve of the model in this figure is AUC = 0.810, the standard error is 0.333, and the 95% confidence interval is [0.764, 0.744], indicating that the discrimination of the established model is good.

[0097] (3) Calibration: It is a meaningful indicator of the accuracy of predicting the probability of a positive event occurring in an individual, and is mainly used to evaluate the consistency of the risk prediction model. The present invention presents this by drawing a calibration graph, see Figure 6 The Hosmer-Lemeshow test was used to evaluate the calibration of the model. The model χ2 was 5.540, P = 0.7849. The results showed that the predicted results of the constructed model were consistent with the actual results.

[0098] (4) Clinical applicability: The clinical applicability of the model was evaluated by drawing a decision curve analysis (DCA). Figure 7 The gray curve represents the clinical net benefit with a negative slope, and the black solid line represents that all patients in the constructed risk prediction model have normal nutrition and the clinical net benefit is 0. The decision curve DCA dotted lines constructed by the present invention are higher than the two extreme lines, indicating that the constructed risk prediction model has certain clinical significance.

[0099] Example 3: Comparison of the Validation Set Risk Prediction Model with the Training Set Risk Prediction Model

[0100] The validation set included 619 elderly patients with COPD who met the study criteria and were admitted to the same hospital between January 2021 and December 2022, see Tables 6 and 7.

[0101] Table 6 Clinical data of elderly patients with COPD Univariate analyses

[0102]

[0103]

[0104] Table 7 Clinical data of elderly patients with COPD Univariate analyses

[0105]

[0106]

[0107] According to the method of establishing the risk prediction model of the training set, the risk prediction model of the validation set is established, and then the risk prediction model of the validation set is evaluated, as follows:

[0108] (1) Discrimination:

[0109] Substitute the validation set data into the risk prediction model to draw the ROC curve, and evaluate the discrimination of the validation model using the area under the ROC curve (AUC). Figure 8 . The area under the ROC curve of the validation set model was AUC = 0.632, the standard error was 0.262, and the 95% confidence interval was [0.700, 0.552]. The results showed that the area under the ROC curve of the risk prediction model constructed by the present invention was tested by statistics (Delong test), and the evaluation of the training set model and the validation set model showed: z = 2.669, P = 0.009 < 0.05, which showed that there was a statistical difference between the training set and the validation set, suggesting that the training set model was significantly better than the validation set model, indicating that the risk prediction model established by the present invention using a training set with more cases is better than the validation set with fewer cases.

[0110] (2) Calibration:

[0111] To evaluate the calibration of the model with the same training set, draw a calibration graph in the software, see Figure 9 The Hosmer-Lemeshow test (goodness of fit test) showed that the model χ 2 =28.646, P=0.0007<0.05. The results showed that there were statistical differences between the observed values ​​and the actual values ​​in the training set and the validation set, and the calibration degree of the risk prediction model constructed in the training set was better than that in the validation set.

[0112] (3) Clinical applicability

[0113] In the decision survival curve analysis, the diagonal line represents that all patients were malnourished and all patients received intervention measures. The horizontal line represents that all patients were well-nourished and no intervention measures were taken, and the net benefit was 0. The net benefit is a backslash with a negative slope, see Figure 10 The figure shows that the decision survival curve is higher than the two extreme lines, which has certain clinical significance. However, compared with the training set decision survival curve, the area under the curve is significantly smaller, indicating that the training set survival decision curve is better than the validation set.

[0114] The risk prediction model constructed by the present invention uses malnutrition as the outcome indicator and substitutes the validation set data into the model's mathematical formula to calculate the risk of malnutrition. The training set prediction model achieved an accuracy of 77.2%, while the validation set prediction model achieved an accuracy of 67.4%. (See Table 8) The training set prediction model outperformed the validation set prediction model. The present invention further validated the model's predictive effectiveness using machine learning.

[0115] Table 8 Validation results of the malnutrition risk prediction model for elderly patients with chronic obstructive pulmonary disease during acute exacerbation

[0116] Tab.8Validation results of risk prediction model for malnutrition inacute exacerbation in

[0117] elderly patients with chronic obstructive pulmonary disease

[0118]

[0119] Example 4: Machine Learning Method to Validate the Training Set Risk Prediction Model

[0120] The present invention uses four algorithms, linear regression, gradient boosting, support vector machine and random forest, to build a model and accurately measure the quality of the malnutrition risk prediction model for elderly patients with COPD during acute exacerbation. Figure 11 and Figure 12 , the linear regression model has a precision of 79.60%, an accuracy of 75.60%, and a recall of 86.70%. Figure 13 and Figure 14 , the gradient boosting machine model has a precision of 81.50%, an accuracy of 81.50%, and a recall of 98.90%. Figure 15 、 Figure 16 、 Figure 17 and Figure 18 ,The precision of the support vector machine model is 75.80%, 76.20%, 72.10%, 70.90%, the accuracy is 73.10%, 73.10%, 70.60%, 68.10%, and the recall is 87.90%, 86.90%, 91.60%, 88.80%. Figure 19 and Figure 20 The precision of the random forest model is 81.40%, the accuracy is 80.20%, and the recall is 92.20%, as shown in Table 9. The four methods predict that the precision, accuracy, and recall are all very high, indicating that the constructed model has high credibility and high accuracy.

[0121] Table 9 LR, GBM, SVM and RF internal validation predictive performance evaluation

[0122]

[0123] Comparative Example

[0124] The present invention provides a comparative example to compare the advantages of Lasso regression in screening variables.

[0125] The present invention uses the Lasso regression model to perform preliminary variable screening, and then conducts multi-factor logistic regression analysis and modeling. The Full-Logistic model does not perform Lasso screening and directly performs logistic analysis modeling.

[0126] For the convenience of describing the variable screening and modeling method of the present invention, it is referred to as Lasso-Logistic model. Compared with the full variable modeling method in the prior art, for the convenience of describing this type of model, it is referred to as Full-Logistic model.

[0127] The results of the two prediction models show that the number of independent variables in the Lasso-Logistic model is 12, and the number of independent variables in the Full-Logistic model is 6. Figure 22 The results show that the Lasso-Logistic model has a higher discrimination ability (AUC, 0.810 vs 0.777), see Figure 21 、 Figure 22 The Lasso-Logistic model is more optimal, that is, the modeling method using the Lasso regression analysis variable screening method is better than the full variable inclusion method. Therefore, in the present invention, the risk prediction model includes a total of 12 risk prediction factors.

[0128] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A malnutrition risk prediction model for elderly patients with chronic obstructive pulmonary disease during acute exacerbation, characterized by: The risk prediction model is a multivariate logistic regression model: P = -12.845 + 1.022 × age + 0.339 × ethnicity + 0.077 × education level + 0.683 × gastrointestinal symptoms + 0.035 × height - 0.126 × body mass index + 1.045 × carbon dioxide partial pressure - 0.077 × chloride ion - 0.017 × serum albumin - 0.004 × prealbumin + 0.005 × high-sensitivity C-reactive protein + 0.479 × lymphocyte count.

2. The malnutrition risk prediction model for elderly patients with chronic obstructive pulmonary disease during acute exacerbation according to claim 1, characterized in that: The risk prediction model was expressed using a nomogram, including score, age, ethnicity, education level, presence or absence of gastrointestinal symptoms, height, body mass index, partial pressure of carbon dioxide, chloride ion, serum albumin, prealbumin, high-sensitivity C-reactive protein, lymphocyte count, total score, linear prediction value, and probability of disease.

3. The malnutrition risk prediction model for elderly patients with chronic obstructive pulmonary disease in acute exacerbation period according to claim 1 or 2, characterized in that: The risk prediction model is to make the multivariate logistic regression model into an online web calculator expression; or the risk prediction nomogram is made into an online web calculator expression.

4. A method for constructing a malnutrition risk prediction model for elderly patients with chronic obstructive pulmonary disease during acute exacerbation, characterized in that: From the clinical baseline data of the normal nutrition group and the malnutrition group in the elderly patients with COPD during acute exacerbation, the indicators with statistical differences between the normal nutrition group and the malnutrition group were selected as risk predictors of malnutrition in the elderly patients with COPD during acute exacerbation. The risk prediction variables were subjected to Lasso regression analysis, and 12 variables were screened out from the risk prediction variables through the outcome effect of the Lasso regression path and Lasso cross-validation as risk prediction factors of the risk prediction model; Using the risk prediction factors to construct a multivariate logistic regression model, i.e., a risk prediction model; The multivariate logistic regression model was: P = -12.845 + 1.022 × age + 0.339 × ethnicity + 0.077 × education level + 0.683 × gastrointestinal symptoms + 0.035 × height - 0.126 × body mass index + 1.045 × partial pressure of carbon dioxide - 0.077 × chloride ion - 0.017 × serum albumin - 0.004 × prealbumin + 0.005 × high-sensitivity C-reactive protein + 0.479 × lymphocyte count.

5. The method for constructing a risk prediction model for malnutrition in elderly patients with chronic obstructive pulmonary disease during acute exacerbation according to claim 4, characterized in that: The risk prediction model is expressed by a nomogram, which includes: score, age, ethnicity, education level, presence or absence of gastrointestinal symptoms, height, body mass index, carbon dioxide partial pressure, chloride ion, serum albumin, prealbumin, high-sensitivity C-reactive protein, lymphocyte count, total score, linear prediction value and risk probability of disease.

6. The method for constructing a malnutrition risk prediction model for elderly patients with chronic obstructive pulmonary disease in the acute exacerbation phase according to claim 4 or 5, characterized in that: After the risk prediction model is established, the multivariate logistic regression model is made into an online web calculator expression, or the risk prediction nomogram is made into an online web calculator expression.

7. The method for constructing a risk prediction model for malnutrition in elderly patients with chronic obstructive pulmonary disease during acute exacerbation according to claim 6, characterized in that: The elderly patients with acute exacerbation of COPD who meet the research criteria are: ① aged ≥ 60 years; ② meet the diagnostic criteria for COPD in the 2023 edition of the "Global Initiative for Chronic Obstructive Pulmonary Disease: Global Strategy for the Diagnosis, Treatment and Prevention of COPD"; ③ symptoms such as cough, sputum, chest tightness, and asthma have worsened, or are accompanied by acute infection symptoms such as fever, and require a change in medication regimen; ④ are conscious and able to communicate normally; ⑤ give informed consent and agree to cooperate with the investigation.

8. The method for constructing a risk prediction model for malnutrition in elderly patients with chronic obstructive pulmonary disease during acute exacerbation according to claim 6, characterized in that: The exclusion criteria for the study were as follows: 1) elderly patients with COPD in the acute exacerbation stage were diagnosed with other lung diseases such as lung cancer, bronchial asthma, interstitial lung disease, etc.; 2) patients with combined heart failure or heart insufficiency; 3) patients with combined metabolic diseases such as hyperthyroidism, hypothyroidism, and diabetes; 4) patients with digestive system diseases or nutritional intervention treatment within 3 months before hospitalization; 5) patients with a history of mental illness.

9. The method for constructing a risk prediction model for malnutrition in elderly patients with chronic obstructive pulmonary disease during acute exacerbation according to claim 6, characterized in that: Among the indicators selected to have statistical differences between the normal nutrition group and the malnutrition group, the statistical differences refer to P≤0.

001.

10. The method for constructing a risk prediction model for malnutrition in elderly patients with chronic obstructive pulmonary disease during acute exacerbation according to claim 6, characterized in that: After the risk prediction model was established, it was verified by machine learning methods, including linear regression, gradient boosting machine, support vector machine and random forest algorithm.

Images