A method and system for early kidney injury risk assessment based on routine physical examination indicators
By using routine physical examination indicators and logistic regression models, the equipment and cost issues of early kidney injury screening in primary communities have been resolved, achieving low-cost and efficient early kidney injury risk assessment, which is applicable to early kidney injury screening in various scenarios.
Patent Information
- Application Number
- CN202611252297.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-18
- Publication Date
- 2026-11-10
AI Technical Summary
Large-scale early kidney injury screening is difficult to implement in grassroots communities. Existing technologies rely on specialized equipment and complex procedures, which are costly and result in a high rate of missed diagnoses of early kidney injury patients.
Using four routine physical examination indicators—age, blood urea nitrogen, uric acid, and urine protein—and combining a logistic regression model and Platt calibration formula, this tool provides a visual nomogram assessment tool suitable for early kidney injury risk assessment in various scenarios.
It enables low-cost, comprehensive screening of early kidney injury populations, with high predictive performance and cross-population adaptability. It is suitable for large-scale automated screening and individual self-examination, has strong clinical interpretability, and can flexibly adapt to the needs of different populations.
Smart Images

Figure CN123072718A_ABST
Abstract
Description
Technical Field
[0001] This application relates to a method and system for early kidney injury risk assessment based on routine physical examination indicators, belonging to the field of medical data processing and chronic disease auxiliary risk assessment technology. Background Technology
[0002] Currently, the clinically recognized gold standard for diagnosing chronic kidney disease (CKD), based on the KDIGO 2024 Clinical Practice Guidelines for Chronic Kidney Disease, requires meeting one of the following criteria for a duration of ≥3 months:
[0003] 1) Estimated glomerular filtration rate (eGFR) <60 mL / min / 1.73m²; 2) Urine albumin / creatinine ratio (ACR) ≥ 30 mg / g; 3) Other abnormal markers of kidney injury, including: abnormal urine sediment (such as red blood cell casts, white blood cell casts, etc.), abnormal imaging examinations (such as changes in kidney morphology and structure, renal cysts, hydronephrosis, etc.), abnormal markers related to renal tubular diseases, and abnormal renal histopathology.
[0004] In the early stage of kidney injury (i.e., when eGFR decreases slightly to 60-90 mL / min / 1.73m² (G2 stage) or ACR increases slightly to 30-299 mg / g (A2 stage), although the diagnostic criteria for CKD have not yet been fully met, it indicates that reversible kidney function damage has begun, which is a critical window for intervention. Screening at this stage also relies on the above indicators.
[0005] However, at the grassroots level, such as community health service centers and township health centers, this screening model faces the following practical difficulties: ACR testing requires specific equipment and reagents, which many grassroots institutions cannot perform. Microscopic examination of urine sediment relies on experienced laboratory personnel, making it difficult to standardize its implementation at the grassroots level. Imaging examinations (such as renal ultrasound) require specialized equipment and ultrasound physicians, making them difficult to popularize in large-scale screening. The aforementioned testing methods are costly and time-consuming, making them difficult to cover in large-scale physical examinations involving thousands of people. The awareness rate of CKD among the grassroots population is extremely low (approximately 8.7%), and a large number of early patients are missed.
[0006] Therefore, there is an urgent need for an early kidney injury risk screening tool that relies solely on routine physical examination indicators, is low-cost, and easy to operate, in order to quickly identify high-risk individuals in large-scale populations at the grassroots level. Summary of the Invention
[0007] The purpose of this invention is to address the shortcomings of existing early kidney injury screening technologies, which rely on specialized equipment (such as ACR detectors and ultrasound equipment), have high testing costs, are complex to operate, and are difficult to implement in large-scale populations in grassroots communities. This application aims to provide an early kidney injury risk assessment method and system that uses only four routine physical examination indicators: age, blood urea nitrogen, uric acid, and urine protein. It is equipped with a logistic regression model, an automated computer assessment process, and a visual nomogram assessment tool. It takes into account both large-scale information-based physical examinations and scenarios where there is no equipment in grassroots communities or home self-examination scenarios. The internal and external independent datasets validate and distinguish excellent performance, enabling low-cost and comprehensive screening of early kidney injury populations in the community.
[0008] To achieve the above objectives, this application adopts the following technical solution: In a first aspect, this application provides a method for assessing the risk of early kidney injury, comprising the following steps: A1. Indicator Input: Collect four indicators for the individual or population to be tested: age, blood urea nitrogen, uric acid, and qualitative urine protein. A2. Data Preprocessing: Standardized Z-values were calculated for age, blood urea nitrogen, and uric acid using a standardized formula. The qualitative results of urine protein were assigned codes according to the coding rules: negative = 0, weakly positive = 1, 1+=2, 2+=3, 3+=4, 4+=5; the standardized formula is: Z=(X... μ) / σ, where X is the original measured value of the indicator, μ is the predetermined mean, and σ is the predetermined standard deviation; A3. Risk Probability Calculation: Substitute the values into the logistic regression formula to calculate the Logit value, and then use the Platt calibration formula to calculate the individual's early kidney injury calibrated risk probability P. The Logit value is calculated as follows: Logit = 0.0086 + 0.4026 × Z_age + 0.3270 × Z_BUN + 0.3735 × Z_UA + 0.1586 × X_urinary protein; where Z_age, Z_BUN, and Z_UA are the standardized values of the corresponding indicators, and X_ is the coded value of urinary protein. The Platt calibration formula is: The calibrated probability P = 1 / [1 + exp( (1.4419×Logit 0.0929))]; A4. Risk Classification Determination: Compare the risk probability P with the predetermined optimal classification threshold P0. If P > P0, the assessment conclusion is: high risk of early kidney injury, suggesting that a detailed specialist examination should be completed at the nephrology department. If P≤P0, output the assessment conclusion: low risk, and suggest regular follow-up examinations according to the routine physical examination cycle.
[0009] Furthermore, this method can be deployed in any of the following: electronic medical record system, physical examination management software, web page, smartphone APP, cloud API, or embedded physical examination device, to achieve automated screening of large numbers of people.
[0010] Furthermore, the mean and standard deviation in A2 are determined by pre-collected samples or physical examination data of the target population; preferably, the mean and variance are: mean age μ=74.5, standard deviation σ=6.8; mean blood urea nitrogen (BUN) μ=5.75, standard deviation σ=1.35; mean uric acid (UA) μ=328, standard deviation σ=78.
[0011] Furthermore, the optimal classification threshold P0 in A4 is determined by pre-collected samples or physical examination data of the target population; preferably, the optimal classification threshold P0 is 0.4797.
[0012] Furthermore, the logistic regression formula in A3 is displayed in a nomogram. Accordingly, A2+A3 is replaced by: reading the individual scores of each indicator based on the actual test values of the four indicators of the individual to be tested on the nomogram; summing the four individual scores to obtain the total score; and reading the corresponding calibrated risk probability on the probability axis of the nomogram based on the total score.
[0013] The nomogram supports three display formats: paper printout, electronic image, and interactive digital version, making it suitable for home self-testing and rapid in-clinic assessment by primary care physicians without electronic devices.
[0014] Secondly, this application provides an early kidney injury risk assessment system, comprising: The indicator acquisition module is used to input or import in batches the four indicators of age, blood urea nitrogen, uric acid, and urinary protein of the individual or population to be tested. Data preprocessing module: Stores the standardized parameters and urine protein coding rules calibrated in claim 1, and completes the standardization of continuous indicators and the coding of qualitative indicators; Probability calculation module: Built-in logistic regression formula and Platt calibration formula for calculating Logit values, outputting the individual risk probability after calibration; Risk assessment module: Stores the baseline threshold of 0.4797 and the threshold adjustment configuration logic, and outputs high / low risk classification conclusions and health guidance suggestions.
[0015] Furthermore, the system platform can be any one of the following: local computer software, cloud-based web platform, mobile app, embedded program in integrated physical examination machine, or API call from a third-party system. Furthermore, the system also includes: Visualization output module: Used to output automated text evaluation reports or a visual interface for manually viewed line graphs; And / or, storage module: stores the model's calibration parameters, population physical examination data, and historical evaluation results.
[0016] Thirdly, this application provides the use of reagents for detecting the concentrations of blood urea nitrogen and uric acid in subjects, as well as reagents for qualitative detection of urinary protein, in the preparation of products for early kidney injury risk screening.
[0017] Furthermore, the product is a reagent or kit, and the kit also includes an instruction manual, which describes the following formula for calculating the risk probability of Platt calibration and its usage instructions: The formula for calculating the probability of risk is P = 1 / [1 + exp( (1.4419×Logit 0.0929))]; where Z is the Logit value calculated based on the logistic regression formula, and the formula for calculating the Logit value is: Logit = 0.0086 + 0.4026×Z_age + 0.3270×Z_BUN + 0.3735×Z_UA +0.1586×X_urinary protein; where Z_age, Z_BUN, and Z_UA are the standardized values of the corresponding indicators, and X_ is the coded value of urinary protein.
[0018] Furthermore, the product is a reagent or kit, and the kit also includes an instruction manual, which describes a nomogram generated based on a logistic regression formula and its usage instructions. The logistic regression formula is: Logit = 0.0086 + 0.4026×Z_age + 0.3270×Z_BUN + 0.3735×Z_UA +0.1586×X_urinary protein; where Z_age, Z_BUN, and Z_UA are the standardized values of the corresponding indicators, and X_ is the coded value of urinary protein.
[0019] Compared with the prior art, this application has the following beneficial effects: 1. Extremely low implementation cost and high accessibility This application relies solely on four routine physical examination indicators: age, blood urea nitrogen, uric acid, and urine protein. These indicators are mandatory items in community health checkups, with extremely low testing costs and 100% equipment availability, requiring no additional equipment investment or special reagents. Compared to existing gold standard solutions that rely on ACR testing (requiring specific equipment), this application can be seamlessly integrated into existing community health checkup processes, enabling large-scale, low-cost screening for early kidney damage, directly solving the technical problems of existing technologies being "highly costly and difficult to implement on a large scale."
[0020] 2. Excellent predictive performance and cross-population generalization ability This application employs rigorous LASSO regression combined with the 3-SE rule to select core features, followed by Platt probability calibration. The constructed logistic regression model achieves an AUC of 0.7611 on the independent internal validation set and 0.8227 on the external validation set. Notably, the external validation set and training set are from different sources (different communities) and have different population characteristics; however, the model's AUC increases instead of decreasing, demonstrating the application's excellent cross-population generalization ability. Calibration and decision curve analysis further confirms a high degree of consistency between the model's predicted probabilities and actual observed probabilities, with significant clinical net benefit. These technical effects are directly attributable to the feature selection method (LASSO + 3-SE rule) and the probability calibration method (Platt calibration) of this application.
[0021] 3. Easy to operate and adaptable to various application scenarios This application provides two parallel implementation schemes, each adapted to different scenarios: Method 1 (System Integration): The model is embedded into the electronic medical record or physical examination system. Users only need to input four indicators, and the computer automatically completes the standardization, calculation, calibration, comparison and output. No manual intervention is required throughout the process, which is suitable for large-scale automated screening.
[0022] Method 2 (Manual Evaluation with Nodal Chart): The model parameters are converted into a visual nodal chart. Users can complete the evaluation in three steps: "drawing lines, adding scores, and reading probabilities". No electronic devices or calculation tools are required. It is suitable for scenarios such as personal self-examination and manual evaluation by primary care physicians.
[0023] Both of the above methods are directly generated by the core parameters (standardization parameters, regression coefficients, calibration parameters, thresholds) and nomogram generation method of this application, achieving the technical effect of "one model, two uses".
[0024] 4. The model has strong interpretability and high clinical reliability. This application employs a logistic regression model, clearly defining the risk contribution direction (positive / negative) and intensity (regression coefficient) of each predictor. SHAP analysis further quantifies the importance ranking of each feature (age contributes the most, followed by uric acid, blood urea nitrogen, and urinary protein) and its direction of action (all are positive risk factors, with urinary protein exhibiting high specificity), which is highly consistent with the pathophysiological mechanisms of early kidney injury. Good interpretability enhances clinicians' confidence in the model's predictive results, facilitating its clinical implementation and widespread adoption.
[0025] 5. Flexible clinical adaptability The classification threshold (0.4797) provided in this application can be flexibly adjusted according to the expected incidence rate of the target population: the threshold can be appropriately lowered (e.g., adjusted to 0.4, 0.3, or lower) for high-incidence populations to improve sensitivity, while the threshold can be appropriately raised (e.g., adjusted to 0.5, 0.6, or higher) for low-incidence populations to improve specificity. Simultaneously, the standardized parameters can be updated based on the distribution of physical examination data from different regions to adapt to local population characteristics. This flexibility allows this application to adapt to the needs of different regions, different populations, and different clinical scenarios, demonstrating promising prospects for widespread application. This effect is directly resulting from the threshold adjustment mechanism and standardized parameter update mechanism set in this application. Attached Figure Description
[0026] Figure 1 is a flowchart of the construction process of the evaluation model of this application; Figure 2 is a flowchart of the automated computer evaluation method of this application; Figure 3 shows the ROC curves of the prediction model of this application on the internal validation set and the external validation set; Figure 4 shows the calibration curves of the prediction model of this application on the internal validation set and the external validation set; Figure 5 shows the decision curve analysis (DCA) of the prediction model of this application on the internal validation set and the external validation set; Figure 6 is a bar chart showing the importance of SHAP features for each feature in the prediction model of this application; Figure 7 is a nomogram of early kidney injury predictions used for manual assessment in this application. Detailed Implementation
[0027] To make the technical solution of this application clearer and easier to understand, preferred embodiments are described in detail below with reference to the accompanying drawings.
[0028] Example 1: Construction of a Risk Assessment Model like Figure 1 As shown, the construction of a risk assessment model specifically includes the following steps: 1. Data Acquisition and Outcome Definition 1) Data sources, inclusion and exclusion criteria, and sample size Both datasets are from community health check centers in Shanghai.
[0029] The Friendship Community initially enrolled 5154 cases. Inclusion criteria: (1) age ≥ 18 years; (2) urinalysis or urinalysis results were available, or although some results were missing, they met the outcome determination rules and could be included. Exclusion criteria were based on the completeness of outcome variables (eGFR and ACR): those with both eGFR and ACR missing were excluded; those with one normal result and the other missing were excluded; those with one abnormal result and the other missing were classified into the high-risk group according to the "abnormal is high-risk" principle. After screening according to the above criteria, a total of 4560 cases were finally included, including 2405 cases in the normal group and 2155 cases in the high-risk group.
[0030] After screening using the same inclusion and exclusion criteria, 1291 cases were ultimately included in Jiangchuan Community.
[0031] 2) Division of training set, internal validation set and external validation set The 4,560 cases from Youyi Community were stratified and sampled at an 8:2 ratio, resulting in a training set of 3,648 cases and an internal validation set of 912 cases; the 1,291 cases from Jiangchuan Community were used as the external validation set.
[0032] The internal validation set and training set originate from the same community. The internal validation set is used for performance testing after the model is developed, and it does not participate in any modeling process such as feature selection, parameter tuning, or threshold determination. The external validation set originates from an independent community and is used separately after the final model is finalized to evaluate the model's cross-population generalization ability.
[0033] 3) Definition of outcome variables According to the KDIGO 2024 Clinical Practice Guideline for Chronic Kidney Disease, the following outcomes are defined: High-risk group for early kidney injury: estimated glomerular filtration rate (eGFR) <90 mL / min / 1.73m², or urine albumin / creatinine ratio (ACR) ≥ 30 mg / g; Normal group: eGFR ≥ 90 mL / min / 1.73m² and ACR < 30 mg / g.
[0034] 2. Core Indicator Screening 1) Missing data screening and initial clinical screening: The missing data rate of all original variables (all original variables are routine physical examination indicators, specifically including 7 categories: physical examination, complete blood count, complete urine test, stool, biochemical test, immune and tumor markers, infection screening, imaging, functional and specialist examinations; a total of 234 indicators) was calculated, and variables with a missing data rate >30% were removed (a total of 72 variables were removed); in order to balance the simplicity of the model and clinical interpretability, 55 variables closely related to renal function, early renal injury and cardiovascular metabolic risk were selected based on clinical significance and entered into subsequent univariate analysis; 2) Univariate analysis (P>0.2 elimination): Perform univariate tests on all remaining variables against the outcome. For continuous variables, use the Welch t test or Mann-Whitney U test; for categorical variables, use the chi-square test; retain variables with P ≤0.2 and eliminate variables with P>0.2. After this step, 37 variables are retained. 3) Multicollinearity test (VIF>10 elimination): The variance inflation factor (VIF) is used to perform a stepwise multicollinearity test on the remaining variables; highly collinear variables with VIF>10 are eliminated stepwise; the specific steps are as follows: ① Calculate the VIF for each of the current candidate variables; ② Identify the variables with VIF > 10, and delete only the one with the largest VIF. ③ Recalculate VIF for all remaining variables; ④ Repeat the loop until all variables VIF ≤ 10; No variables were removed in this step, and the number of candidate variables remains at 37. 4) Pearson correlation screening (redundancy removal if |r|>0.8): For the variable set after VIF screening, calculate the pairwise Pearson correlation coefficient matrix between variables. When a pair of variables (|r|>0.8) is considered highly redundant, retain the variable with the higher correlation to the outcome. In this step, red blood cell count (RBC) was found to be highly correlated with hemoglobin (HGB), so RBC was removed and HGB was retained. The number of candidate predictive variables was reduced from 37 to 36. 5) Feature selection was performed using LASSO regression combined with the 3-times-standard-error (3-SE) rule: The optimal regularization parameter α was selected based on 10-fold cross-validation. By comparing the 1-SE, 2-SE, 2.5-SE, and 3-SE rules, the 3-SE criterion was finally adopted to determine α=0.032237, and four core features with non-zero coefficients were selected. Compared with the model corresponding to the minimum error α value, the coefficient of determination R² under the 3-SE rule decreased from 0.1827 to 0.1471, a decrease of only 0.0355, with no substantial loss in predictive performance. However, the number of features was greatly reduced from 36 to 4, which greatly improved the model's generalization ability and clinical applicability. 6) 1000 Bootstrap Stability Analysis: To evaluate the stability of features selected by LASSO and avoid false positives from random sampling, the dataset was sampled with replacement to generate 1000 Bootstrap resampled datasets. The screening steps 1)-5) were repeated for each resampled sample. The inclusion frequency (number of occurrences / 1000) of each variable in the 1000 resamples was calculated. Features with an inclusion frequency ≥70% were considered stable, and those below 70% were discarded.
[0035] After the above screening steps, age, blood urea nitrogen (BUN), uric acid (UA), and urinary protein were finally determined as the four core indicators.
[0036] 3. Model Training and Selection Using four core metrics as input variables, six machine learning models were constructed: Logistic Regression (LR), Random Forest (RF), XGBoost, Support Vector Machine (SVM), Naive Bayes (NB), and Neural Network (NN). All models underwent 10-fold hierarchical cross-validation on the training set (3648 examples). Results showed that the Logistic Regression model had the highest AUC (0.7337) and the lowest Brier score (0.2063), and also exhibited the simplest model structure and the strongest interpretability. Therefore, Logistic Regression was chosen as the final model.
[0037] 4. Model parameter determination Determine the parameters of the logistic regression model on the training set: (1) Standardization parameter (used for standardization of continuous variables: Z = (X-μ) / σ)
[0038] (2) Urine protein coding rules
[0039] (3) Logistic regression coefficient Logit = 0.0086 + 0.4026 × Z_age + 0.3270 × Z_BUN + 0.3735 × Z_UA + 0.1586 × X_urine protein (4) Platt calibration parameters Perform Platt calibration on the model output: After calibration, the probability P = 1 / [1 + exp( (1.4419×Logit-0.0929))] (5) Classification threshold Using 10-fold cross-validation to maximize the Youden index, a classification threshold of 0.4797 was determined. A value greater than 0.4797 was considered high-risk, and a value less than or equal to 0.4797 was considered low-risk.
[0040] 5. Model Validation The model was validated using an independent internal validation set (912 cases) and an external validation set (1291 cases). The ROC curves are shown below. Figure 3 As shown, the internal validation set AUC = 0.7611 and the external validation set AUC = 0.8227.
[0041] Calibration curve as follows Figure 4 As shown, the model's predicted probabilities are largely consistent with the actual observed probabilities. Decision curve analysis (DCA) is as follows... Figure 5 As shown, within the clinical threshold range of 0.10–0.55, the model’s net benefit is significantly better than the extreme strategies of “all treatment” and “no treatment”.
[0042] 6. Model Interpretability Analysis (SHAP) The SHAP method was used to analyze the contribution of each feature to the prediction results: Feature importance ranking as follows Figure 6 As shown, age contributes the most, followed by uric acid, blood urea nitrogen, and urinary protein.
[0043] Direction of action: All four indicators are positive risk factors, with higher values indicating greater risk. Among them, proteinuria only shows a significant risk indicator effect when positive, exhibiting high specificity.
[0044] 7. Generating Column Charts The nomogram corresponding to the above logistic regression model was generated using Python and the Matplotlib plotting library, as shown below. Figure 7 As shown. Users do not need to perform formula calculations; they can obtain individualized risk probabilities through three steps: "drawing lines - adding points - reading probabilities".
[0045] II. Model Usage Method This application provides two parallel model usage methods, each suitable for different scenarios: Method 1 is suitable for medical institutions equipped with electronic systems for large-scale screening; Method 2 uses the nomogram as an independent tool for individual self-examination, family self-testing, or manual assessment by primary care physicians.
[0046] Method 1: Automated Evaluation Method The logistic regression model and parameters of this application are deployed on computer devices (such as electronic medical record systems, physical examination information management systems, personal computers, or smartphones), and the computer automatically executes the calculation steps, such as... Figure 2 As shown.
[0047] 1. Input four indicators Obtain the individual's age, blood urea nitrogen (BUN), uric acid (UA), and urine protein (qualitative results).
[0048] 2. Computers automatically calculate risk probabilities. The computer performs the following calculations in the background in sequence: (1) Standardization: Age, BUN, and UA are standardized according to the following formula: Z = (X) μ) / σ Where μ and σ are the standardized parameters determined in the modeling steps above; age: μ=74.5, σ=6.8; BUN: μ=5.75, σ=1.35; UA: μ=328, σ=78.
[0049] (2) Urine protein coding: The qualitative results of urine protein are coded according to the following rules: negative = 0, weak positive = 1, 1+ = 2, 2+ = 3, 3+ = 4, 4+ = 5.
[0050] (3) Calculate the Logit value: Substitute it into the logistic regression equation: Logit = 0.0086 + 0.4026 × Z_age + 0.3270 × Z_BUN + 0.3735 × Z_UA + 0.1586 × X_urine protein (4) Platt calibration: Converting Logit values into calibrated risk probabilities: Probability after calibration = 1 / (1+exp( (1.4419×Logit-0.0929))) 3. Output evaluation results The calculated risk probability P is compared with the threshold of 0.4797: If P>0.4797, output "High risk of early kidney injury" and suggest "Further specialist examination recommended"; If P ≤ 0.4797, output "Low risk" and suggest "It is recommended to have a regular follow-up examination".
[0051] Method 2: Manual evaluation based on nomogram For small-scale or scattered assessment scenarios such as individual home self-assessment, family health management, rapid assessment in primary care physician clinics, and remote areas without electronic systems, technologies such as... can be used... Figure 7 The nomogram shown is used for evaluation.
[0052] 1. Obtain individual scores by examining images. On the nomogram, based on the individual's age, blood urea nitrogen, uric acid, and urinary protein levels, the corresponding individual scores (Points) are found.
[0053] 2. Calculate the total score The four individual scores are added together to obtain the total score.
[0054] 3. Read the risk probability Find the risk probability value corresponding to the total score at the bottom of the nomogram.
[0055] 4. Determine the risk level The probability of reading risk is compared with the threshold of 0.4797 marked on the nomogram: If the risk probability is >0.4797, it is considered a high risk of early kidney injury; If the risk probability is ≤ 0.4797, it is judged as low risk.
[0056] Application Example 1: Automated Computer Assessment - Low-Risk Individuals Subject to be tested: A 65-year-old male with the following physical examination indicators: age 65, BUN 6.5mmol / L, UA 380μmol / L, and negative for urine protein.
[0057] 1. Standardized calculation: Z_age=(65) 74.5) / 6.8≈ 1.40 Z_BUN=(6.5 5.75) / 1.35≈0.56 Z_UA=(380 328) / 78≈0.67 Negative urine protein code: X_urine protein = 0; 2. Logit value calculation: Logit = 0.0086 + 0.4026×(-1.40) + 0.3270×0.56 + 0.3735×0.67 +0.1586×0 ≈ 0.14; 3. Platt calibration probability: P = 1 / [1+exp ( (1.4419×(-0.14) 0.0929))]≈0.45; 4. Risk assessment: 0.45 < 0.4797, output assessment result: low risk of early kidney damage, it is recommended to have a routine physical examination to check kidney function every year.
[0058] Application Example 2: Automated Computer Assessment - High-Risk Individuals Subject to testing: 78-year-old female, physical examination indicators: age 78, BUN 8.2mmol / L, UA 420μmol / L, urine protein 1+.
[0059] 1. Standardized calculation: Z_age=(78) 74.5) / 6.8≈0.51 Z_BUN=(8.2 5.75) / 1.35≈1.81 Z_UA=(420 328) / 78≈1.18 Urine protein 1 + code: X_urine protein = 2; 2. Logit value calculation: Logit=0.0086 + 0.4026×0.51 + 0.3270×1.81 + 0.3735×1.18 + 0.1586×2≈1.35; 3. Platt calibration probability: P=1 / [1+exp( (1.4419×1.35 0.0929))]≈0.85; 4. Risk assessment: 0.85 > 0.4797, output assessment result: high risk of early kidney injury, it is recommended to go to the nephrology department to complete ACR and kidney ultrasound examination.
[0060] Application Example 3: Manual Assessment of Nodal Charts - Borderline High-Risk Individuals Subject to be tested: 72-year-old male, indicators: age 72, BUN 7.1mmol / L, UA 350μmol / L, weak positive for urine protein.
[0061] 1. Scoring by item in the nodal chart: Reference Figure 7 The scores for 72 years old are 13.6, BUN 7.1 corresponds to 13.3, UA 350 corresponds to 12.8, and weak positive urine protein corresponds to 12.6. 2. Total score: 13.6 + 13.3 + 12.8 + 12.6 = 52.3 points; 3. Risk probability of reading: A total score of 52.3 corresponds to a risk probability of 0.63; 4. Judgment: 0.63 > 0.4797, indicating a high risk of early kidney damage, further screening by a specialist is recommended.
[0062] Application Example 4: Evaluation System Web-based Batch Screening Application The application evaluation system was deployed on the community health check cloud web platform. By importing Excel data of 5,000 community elderly health checkups, the system automatically completed the standardization of all samples, probability calculation, risk classification, and batch export of high-risk population lists. Community doctors can then specifically notify high-risk individuals to be referred to the nephrology department, achieving large-scale, low-cost centralized screening.
[0063] Application Example 5: Home Self-Check of Line Drawing Printed Version Print out the line drawing document of this application for community residents to manually assess their health check reports at home. This allows them to independently complete risk self-testing in situations without computers or internet access, and eliminates the need for information technology equipment for grassroots family health follow-ups.
[0064] The above description is merely a preferred embodiment of this application and is not intended to limit this application in any form or substance. It should be noted that those skilled in the art can make several improvements and additions without departing from this application, and these improvements and additions should also be considered within the scope of protection of this application.
Claims
1. A method for assessing the risk of early kidney injury, characterized in that, Including the following steps: A1. Indicator Input: Collect four indicators for the individual or population to be tested: age, blood urea nitrogen, uric acid, and qualitative urine protein. A2. Data Preprocessing: Standardized Z-values were calculated for age, blood urea nitrogen, and uric acid using a standardized formula. The qualitative results of urine protein were assigned codes according to the coding rules: negative = 0, weakly positive = 1, 1+=2, 2+=3, 3+=4, 4+=5; the standardized formula is: Z=(X... μ) / σ, where X is the original measured value of the indicator, μ is the predetermined mean, and σ is the predetermined standard deviation; A3. Risk Probability Calculation: Substitute the values into the logistic regression formula to calculate the Logit value, and then use the Platt calibration formula to calculate the calibrated risk probability P of early kidney injury; where the formula for calculating the Logit value is: Logit = 0.0086 + 0.4026 × Z_age + 0.3270 × Z_BUN + 0.3735 × Z_UA + 0.1586 × X_urinary protein; where Z_age, Z_BUN, and Z_UA are the standardized values of the corresponding indicators, and X_ is the coded value of urinary protein; the Platt calibration formula is: The calibrated probability P = 1 / [1 + exp( (1.4419×Logit 0.0929))]; A4. Risk Classification Determination: Compare the risk probability P with the predetermined optimal classification threshold P0. If P > P0, the assessment conclusion is: high risk of early kidney injury, suggesting that a detailed specialist examination should be completed at the nephrology department. If P≤P0, output the assessment conclusion: low risk, and suggest regular follow-up examinations according to the routine physical examination cycle.
2. The risk assessment method according to claim 1, characterized in that, This method can be deployed in any of the following systems: electronic medical record system, physical examination management software, web page, smartphone APP, cloud API, or physical examination embedded device, to achieve automated screening of large numbers of people.
3. The risk assessment method according to claim 1, characterized in that, The mean and standard deviation in A2 are determined by pre-collected samples or physical examination data of the target population; preferably, the mean and variance are: mean age μ=74.5, standard deviation σ=6.8; mean blood urea nitrogen (BUN) μ=5.75, standard deviation σ=1.35; mean uric acid (UA) μ=328, standard deviation σ=78.
4. The risk assessment method according to claim 1, characterized in that, The optimal classification threshold P0 in A4 is determined by pre-collected samples or physical examination data of the target population; preferably, the optimal classification threshold P0 is 0.4797.
5. The risk assessment method according to claim 1, characterized in that, The logistic regression formula in A3 is presented as a nomogram, which includes four indicator axes: age, blood urea nitrogen, uric acid, and urinary protein, as well as a total score and a probability axis. The four indicator axes are marked with indicator values and corresponding scores. Accordingly, A2+A3 is replaced by: reading the individual scores of each of the four indicators on the nomogram based on the actual test values of the individual; summing the four individual scores to obtain the total score; and reading the corresponding calibrated risk probability on the probability axis of the nomogram based on the total score.
6. An early kidney injury risk assessment system, characterized in that, include: The indicator acquisition module is used to input or import in batches the four indicators of age, blood urea nitrogen, uric acid, and urinary protein of the individual or population to be tested. Data preprocessing module: Stores the standardized parameters and urine protein coding rules calibrated in claim 1, and completes the standardization of continuous indicators and the coding of qualitative indicators; Probability calculation module: Built-in logistic regression formula and Platt calibration formula for calculating Logit values, outputting the individual risk probability after calibration; Risk assessment module: Stores the baseline threshold of 0.4797 and the threshold adjustment configuration logic, and outputs high / low risk classification conclusions and health guidance suggestions.
7. The risk assessment system according to claim 6, characterized in that, The system can be any one of the following: local computer software, cloud web platform, mobile APP, embedded program of integrated physical examination machine, or API call of third-party system. The system also includes: Visualization output module: Used to output automated text evaluation reports or a visual interface for manually viewed line graphs; And / or, storage module: stores the model's calibration parameters, population physical examination data, and historical evaluation results.
8. Application of reagents for detecting blood urea nitrogen and uric acid concentrations in subjects, as well as reagents for qualitative detection of urinary protein, in the preparation of products for early kidney injury risk screening.
9. The application according to claim 8, characterized in that, The product is a reagent or kit, and the kit also includes an instruction manual, which describes the following formula for calculating the risk probability of Platt calibration and its usage instructions: The formula for calculating the probability of risk is P = 1 / [1 + exp( (1.4419×Logit 0.0929))]; where Z is the Logit value calculated based on the logistic regression formula, and the formula for calculating the Logit value is: Logit = 0.0086 + 0.4026×Z_age + 0.3270×Z_BUN + 0.3735×Z_UA + 0.1586×X_urinary protein; where Z_age, Z_BUN, and Z_UA are the standardized values of the corresponding indicators, and X_ is the urinary protein code value.
10. The application according to claim 8, characterized in that, The product is a reagent or kit, and the kit also includes an instruction manual. The instruction manual describes a nomogram generated based on a logistic regression formula and its usage instructions. The logistic regression formula is: Logit = 0.0086 + 0.4026×Z_age + 0.3270×Z_BUN + 0.3735×Z_UA + 0.1586×X_urinary protein; where Z_age, Z_BUN, and Z_UA are the standardized values of the corresponding indicators, and X_ is the urinary protein code value.
