Rehospitalization risk assessment method and system based on heterogeneous electronic medical record data adaptation
Patent Information
- Application Number
- CN202611311336.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-27
- Publication Date
- 2026-10-27
AI Technical Summary
以解决不同医疗机构电子病历数据字段及检验指标报告形式存在差异，导致既有再入院风险评估模型跨机构应用时需要额外数据映射或模型适配，以及部分现有模型输入变量较多、计算过程较复杂、难以实现简便风险分层的问题
[0027]1.本发明通过设置红细胞分布宽度相关检验字段的识别与替代判定逻辑，使系统能够自动识别不同电子病历系统中的不同检验字段，并在红细胞分布宽度标准差字段不可用时自动切换至红细胞分布宽度变异系数字段继续执行判定，避免了因字段缺失导致的数据处理流程中断，提高了风险评估方法在异构电子病历环境下的可实施性，减少了跨机构数据适配所需的额外数据整理与字段映射操作。
Smart Images
Figure FT_1 
Figure FT_2
Abstract
Description
Technical Field
[0001] This invention belongs to the field of medical and health information technology and data processing technology, specifically relating to a readmission risk assessment method and system based on heterogeneous electronic medical record data adaptation. Background Technology
[0002] With the development of electronic medical record systems and medical information technology, early assessment of the risk of unplanned all-cause readmission within 30 days after discharge, utilizing demographic information, disease diagnosis, hospitalization process, and laboratory test fields generated during patient hospitalization, has become an important application area for clinical risk management and medical data analysis. Current technologies typically use logistic regression, machine learning, or other statistical modeling methods to screen clinical features related to readmission from electronic medical record data, establish risk prediction models, and assess or warn of patient readmission risk based on the model calculation results.
[0003] However, existing readmission risk assessment technologies still have certain limitations. First, some predictive models rely on a large number of clinical variables or complex model calculation processes, and have high requirements for the integrity of electronic medical record data. When some key variables are missing, the model may not be able to be calculated directly or its predictive performance may decline, which is not conducive to rapid deployment and application in different medical institutions.
[0004] Secondly, electronic medical records and laboratory information systems across different medical institutions vary in data field settings, indicator names, and reporting methods. The same clinical information may be presented using different data fields or formats. For example, red blood cell distribution width (RDW) can be reported using either RDW-SD or RDW-CV, and different medical institutions may routinely only provide one of these indicators. If existing risk assessment models rely on a specific field, additional data processing, field mapping, or model adaptation may be required when the target medical institution does not provide that field, thus affecting the cross-institutional applicability and automation level of the risk assessment method.
[0005] Furthermore, existing methods for predicting readmission risk primarily output continuous predicted probabilities or rely on complex models for calculation, lacking standardized scoring and risk stratification mechanisms that can be completed based on limited routine clinical data. In actual discharge management and follow-up scenarios, there is a need for a method that can automatically extract limited key data from electronic medical records, identify and adapt heterogeneous fields, and further complete unified risk calculation and stratified output to reduce data processing complexity and improve feasibility across different medical institutions.
[0006] Therefore, it is necessary to provide a readmission risk assessment technology for heterogeneous electronic medical record data. By identifying, adapting and standardizing key clinical fields, risk scoring and risk stratification can still be completed even when there are differences in data fields in different medical institutions. This provides a simple, stable and easily automated data processing solution for early risk assessment of unplanned all-cause readmissions within 30 days after discharge. Summary of the Invention
[0007] To address the problems existing in the prior art, this invention provides a method and system for readmission risk assessment based on heterogeneous electronic medical record data adaptation. The purpose is to acquire relevant test fields such as the patient's active malignant tumor status, length of hospital stay, and red blood cell distribution width (RDW). When the RDW-SD field is available, an RDW abnormality is determined by a value >45 fL. When the RDW-SD field is missing or unavailable, but the RDW-CV field is available, a value >13.45% is used as a substitute. An active malignant tumor score of 5, a hospital stay >14 days score of 2, and an RDW abnormality score of 1 are set. A comprehensive readmission risk score (ALR score) ranging from 0 to 8 is calculated and divided into low, medium, and high risk levels. This system can adapt to differences in RDW test fields among different medical institutions and can assist in early risk stratification for the first unplanned all-cause readmission within 30 days after discharge. This addresses the issues that differ between electronic medical record data fields and test indicator report formats across different medical institutions, which necessitates additional data mapping or model adaptation when existing readmission risk assessment models are applied across institutions. It also addresses the problems that some existing models have a large number of input variables, complex calculation processes, and difficulty in achieving simple risk stratification.
[0008] To achieve the above objectives, the specific solution of the present invention is as follows:
[0009] A readmission risk assessment method based on heterogeneous electronic medical record data adaptation, wherein the method is executed by a computer system and includes the following steps:
[0010] Step S1, Data Acquisition and Feature Extraction: Obtain the target patient's medical record data during the indexed hospitalization and at discharge from the electronic medical record database, and extract the active malignant tumor status, hospitalization time characteristics, and red blood cell distribution width related test fields from the medical record data;
[0011] Step S2: Field Identification and Adaptation and Status Determination: The cell distribution width related test fields described in Step S1 are searched and replaced. If the red blood cell distribution width standard deviation field is found to be available, the value of the red blood cell distribution width standard deviation field is extracted and compared with 45fL. When the value of the red blood cell distribution width standard deviation field is greater than 45fL, the red blood cell distribution width status is determined to be abnormal; otherwise, the red blood cell distribution width status is determined to be non-abnormal.
[0012] If the erythrocyte distribution width standard deviation field is missing or unavailable, the system automatically switches to the alternative decision path, retrieves the erythrocyte distribution width coefficient of variation field, extracts the value of the erythrocyte distribution width coefficient of variation field, and compares it with 13.45%. If the value of the erythrocyte distribution width coefficient of variation field is greater than 13.45%, the erythrocyte distribution width status is determined to be abnormal; otherwise, the erythrocyte distribution width status is determined to be normal. The 13.45% is an alternative decision threshold pre-calculated based on the principle of maximizing Youden's index, with the target status being a erythrocyte distribution width standard deviation field value greater than 45 fL.
[0013] If both the standard deviation of red blood cell distribution width and the coefficient of variation of red blood cell distribution width are missing or unavailable, the red blood cell distribution width status is determined to be missing data.
[0014] Step S3: Risk Score: Based on the active malignant tumor status and length of hospitalization described in Step S1, and the red blood cell distribution width status determined in Step S2, assign 5 points to the presence of active malignant tumor, 2 points to the length of hospitalization greater than 14 days, and 1 point to the abnormal red blood cell distribution width status. The above three scores are added together to generate an ALR score ranging from 0 to 8 points.
[0015] Step S4, Risk Stratification Output: Based on the comprehensive re-admission risk score described in step S3, and according to the preset mapping relationship between the score range and the risk level, output the corresponding risk level, and output the risk level, the comprehensive re-admission risk score, the active malignant tumor status described in step S1, the length of hospital stay, and the red blood cell distribution width status determined in step S2 to the electronic medical record terminal, the discharge management system, or the follow-up management system.
[0016] Furthermore, the active malignant tumor status mentioned in step S1 is used for identification based on the structured diagnostic fields or preset codes and field mappings in the electronic medical record database; the hospitalization time feature is used to extract the patient's admission time and discharge time, and calculate the actual hospitalization time in days; the red blood cell distribution width related test field is the available field and its value of the laboratory test field in the electronic medical record database.
[0017] Furthermore, the mapping relationship between the preset scoring range and risk level mentioned in step S4 is as follows: ALR scores of 0 to 1 are classified as low risk, scores of 2 to 5 are classified as medium risk, and scores of 6 to 8 are classified as high risk.
[0018] A readmission risk assessment system based on heterogeneous electronic medical record data adaptation includes:
[0019] The data acquisition and preprocessing module is used to acquire the target patient's medical record data during the indexed hospitalization and at discharge through the electronic medical record data interface, and extract the active malignant tumor status, hospitalization time and red blood cell distribution width related test fields from the medical record data;
[0020] The field identification and adaptation module, communicatively connected to the data acquisition and preprocessing module, receives red blood cell distribution width (RBV) related test fields and performs retrieval and substitution determination on these fields. If the RBV standard deviation field is available, its value is extracted and compared with 45 fL. If the value is greater than 45 fL, the RBV status is determined to be abnormal; otherwise, it is determined to be normal. If the RBV standard deviation field is missing or unavailable, the module automatically switches to the substitution determination path and retrieves the RBV variation coefficient. If the data is available, the value of the red blood cell distribution width coefficient of variation field is extracted and compared with 13.45%. If the value of the red blood cell distribution width coefficient of variation field is greater than 13.45%, the red blood cell distribution width status is determined to be abnormal; otherwise, the red blood cell distribution width status is determined to be normal. The 13.45% is an alternative judgment threshold pre-calculated based on the principle of maximizing Youden's index, with the target status being a red blood cell distribution width standard deviation field value greater than 45 fL. If both the red blood cell distribution width standard deviation field and the red blood cell distribution width coefficient of variation field are missing or unavailable, the red blood cell distribution width status is determined to be missing data.
[0021] The risk score calculation module is communicatively connected to the data acquisition and preprocessing module and the field identification and adaptation module, respectively. It is used to assign 5 points to the presence of active malignant tumor, 2 points to the state of hospitalization longer than 14 days, and 1 point to the state of abnormal red blood cell distribution width based on the status of active malignant tumor, length of hospitalization and red blood cell distribution width. The above three scores are added together to generate an ALR score ranging from 0 to 8 points.
[0022] The risk stratification output module is communicatively connected to the risk score calculation module. It is used to output the corresponding risk level according to the ALR score and the preset mapping relationship between the score range and the risk level. The risk level, ALR score, active malignant tumor status, length of hospital stay, and red blood cell distribution width status are output to the electronic medical record terminal, discharge management system, or follow-up management system.
[0023] A computer device includes a processor and a memory, wherein the memory stores a computer program, and when the processor executes the computer program, it implements a readmission risk assessment method.
[0024] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a readmission risk assessment method.
[0025] Advantages of the present invention
[0026] Compared with the prior art, the present invention has the following beneficial effects:
[0027] 1. This invention enables the system to automatically identify different test fields in different electronic medical record systems by setting up identification and substitution judgment logic for red blood cell distribution width related test fields. When the red blood cell distribution width standard deviation field is unavailable, it automatically switches to the red blood cell distribution width coefficient of variation field to continue the judgment. This avoids interruption of the data processing flow due to missing fields, improves the feasibility of risk assessment methods in heterogeneous electronic medical record environments, and reduces the additional data processing and field mapping operations required for cross-institutional data adaptation.
[0028] 2. The data dimensions required by this invention are only three items: active malignant tumor status, hospitalization time, and red blood cell distribution width status. Furthermore, integer scoring is used for ALR score calculation, which reduces the resource consumption required for data acquisition and processing.
[0029] 3. The field retrieval conditions, judgment thresholds, alternative judgment paths, integer scoring, and risk level mapping relationships of this invention are all predetermined calculation logics that can be pre-configured in the system. In practical applications, there is no need to retrain the model or estimate parameters for each patient. The system can automatically execute all steps in the following order: data acquisition, field availability judgment, RDW judgment path selection, ALR score calculation, and risk level output, thereby reducing manual operation.
[0030] 4. This invention outputs the risk assessment results to electronic medical record terminals, discharge management systems, or follow-up management systems, thereby enabling the integration of risk assessment data with existing medical information systems.
[0031] 5. This invention calculates the ALR score by assigning integer scores to the active malignant tumor status, length of hospital stay, and red blood cell distribution width status, and then converts the ALR score into three risk levels (low, medium, and high) according to a preset mapping relationship, thereby achieving standardized output of risk assessment results. Attached Figure Description
[0032] Figure 1 This is a flowchart of the readmission risk assessment method based on heterogeneous electronic medical record data adaptation according to the present invention.
[0033] Figure 2 This is a schematic diagram of the readmission risk assessment system based on heterogeneous electronic medical record data according to the present invention. Detailed Implementation
[0034] The present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. It should be noted that the specific embodiments are not intended to limit the scope of the present invention.
[0035] like Figure 1 As shown in the illustration, this specific embodiment provides a readmission risk assessment method based on heterogeneous electronic medical record data adaptation. The method is executed by a computer system and includes the following steps:
[0036] Step S1, Data Acquisition and Feature Extraction:
[0037] The system retrieves the target patient's medical record data during their indexed hospitalization and discharge from the electronic medical record database, and extracts the active malignant tumor status, hospitalization time characteristics, and red blood cell distribution width related test fields from the medical record data. The active malignant tumor status is used to identify whether the patient has active malignant tumor characteristics based on the structured diagnostic fields or preset codes and field mappings in the electronic medical record database. The hospitalization time characteristics are used to extract the patient's admission time and discharge time to calculate the actual hospitalization time in days. The red blood cell distribution width related test field is the available fields and their values in the laboratory test fields in the electronic medical record database.
[0038] Step S2: Field Identification and Adaptation and Status Determination: The cell distribution width related test fields described in Step S1 are searched and replaced. If the red blood cell distribution width standard deviation (hereinafter referred to as RDW-SD) field is available, the value of the RDW-SD field is extracted and compared with 45fL. When RDW-SD > 45fL, the red blood cell distribution width (hereinafter referred to as RDW) status is determined to be abnormal. When RDW-SD ≤ 45fL, the RDW status is determined to be non-abnormal.
[0039] If the RDW-SD field is missing or unavailable, the system automatically switches to the alternative decision path and retrieves the erythrocyte distribution width coefficient of variation (RDW-CV) field. The value of the RDW-CV field is extracted and compared with 13.45%. If RDW-CV > 13.45%, the RDW status is determined to be abnormal; otherwise, the RDW status is determined to be normal. The 13.45% threshold is an alternative decision threshold pre-calculated using receiver operating characteristic (ROC) analysis and based on the principle of maximizing the Youden index, with RDW-SD > 45fL as the target status. It is not a direct numerical conversion between RDW-SD and RDW-CV.
[0040] If both the RDW-SD and RDW-CV fields are missing or unavailable, the RDW status is determined to be missing data; in step S3, the score corresponding to the red blood cell distribution width status is not calculated.
[0041] Step S3: Risk Scoring Step: Based on the active malignant tumor status and length of hospitalization described in Step S1, and the RDW status determined in Step S2, input the data into the preset risk scoring calculation module, and assign scores according to the following requirements:
[0042] 5 points are awarded if an active malignant tumor is present; otherwise, 0 points are awarded.
[0043] 2 points are awarded if the hospital stay is longer than 14 days; otherwise, 0 points are awarded.
[0044] RDW is assigned 1 point if it is determined to be in an abnormal state, otherwise it is assigned 0 points.
[0045] The scores of the above three items are added together to generate an ALR score ranging from 0 to 8. ALR stands for Activemalignancy, Length of stay, and RDW. ALR is an abbreviation of the first letters of the English names of the three core indicators.
[0046] Step S4, Risk Stratification Output Step: Based on the ALR score described in Step S3, output the corresponding risk level according to the preset mapping relationship between the score range and the risk level. The preset mapping relationship between the score range and the risk level is as follows:
[0047] A score of 0 to 1 indicates low risk;
[0048] A score of 2 to 5 indicates a medium risk level.
[0049] A score of 6 to 8 indicates a high risk level.
[0050] The risk level, ALR score, active malignant tumor status described in step S1, length of hospital stay, and RDW status determined in step S2 are output to the electronic medical record terminal, discharge management system, or follow-up management system to assist in early risk stratification of unplanned all-cause readmissions 30 days after discharge.
[0051] A readmission risk assessment system based on heterogeneous electronic medical record data adaptation, including
[0052] The data acquisition and preprocessing module is used to acquire the target patient's medical record data during the indexed hospitalization and at discharge through the electronic medical record data interface, and extract the active malignant tumor status, hospitalization time and red blood cell distribution width related test fields from the medical record data;
[0053] The field identification and adaptation module, which is communicatively connected to the data acquisition and preprocessing module, is used to receive the erythrocyte distribution width (RDW) related test field and perform retrieval and substitution determination on the RDW related test field: If the RDW-SD field is found to be available, the value of the RDW-SD field is extracted and compared with 45fL. When RDW-SD > 45fL, the RDW status is determined to be abnormal; when RDW-SD ≤ 45fL, the RDW status is determined to be non-abnormal. If the RDW-SD field is missing or unavailable, the system automatically switches to the substitution determination path, retrieves the RDW-CV field, extracts the value of the RDW-CV field and compares it with 13.45%. When RDW-CV > 13.45%, the RDW status is determined to be abnormal; otherwise, the RDW status is determined to be non-abnormal. The 13.45% is a substitution determination threshold obtained in advance through ROC analysis and based on the principle of maximizing the Youden index, with the state of RDW-SD > 45fL as the target state. It is not a direct numerical conversion between RDW-SD and RDW-CV. If both the RDW-SD and RDW-CV fields are missing or unavailable, the RDW status is determined to be missing data.
[0054] The risk score calculation module is communicatively connected to the data acquisition and preprocessing module and the field identification and adaptation module, respectively. It is used to assign 5 points to the presence of active malignancy, 2 points to the state of hospitalization longer than 14 days, and 1 point to the state of abnormal red cell distribution width, based on the status of active malignancy, length of hospitalization, and red cell distribution width. The above three scores are accumulated to generate an ALR score ranging from 0 to 8 points. The ALR score is a readmission risk score constructed based on three core indicators: active malignancy, length of hospitalization, and red cell distribution width (RDW).
[0055] The risk stratification output module is communicatively connected to the risk score calculation module. It is used to output the corresponding risk level according to the ALR score and the preset mapping relationship between the score range and the risk level. The risk level, ALR score, active malignant tumor status, length of hospital stay, and red blood cell distribution width status are output to the electronic medical record terminal, discharge management system, or follow-up management system.
[0056] A computer device includes a processor and a memory, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the above-described readmission risk assessment method.
[0057] A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor executes the computer program to implement the above-mentioned readmission risk assessment method.
[0058] The following are examples used to verify the above method:
[0059] Example 1: Parameter Determination and Risk Assessment Based on RDW-SD Indicators
[0060] This embodiment illustrates the process of determining the scoring parameters of the above method and its application effect in a data environment that typically provides RDW-SD fields.
[0061] Electronic medical record data were obtained from 2,543 discharged patients in the primary study cohort. Restricted cubic spline analysis was used to evaluate the changing trends of continuous variables and 30-day readmission risk, and hospital stay >14 days and RDW-SD >45fL were determined as classification thresholds. Furthermore, the Sullivan method was used, with the minimum regression coefficient β=0.4766 corresponding to RDW-SD >45fL as the base divisor for integer scoring. Finally, an ALR scoring system was established with 5 points for active malignancy, 2 points for hospital stay >14 days, and 1 point for RDW-SD >45fL, with a score range of 0–8 points.
[0062] The electronic medical record data is processed according to the above preset parameters. The active malignant tumor status, length of hospital stay and RDW-SD status of each patient are determined. The ALR score is calculated, and the risk level is divided into low risk, medium risk and high risk according to the score range of 0-1, 2-5 and 6-8 respectively.
[0063] The results showed that the ARROC of the ARR score in the primary study cohort was 0.774; the actual 30-day readmission rates for the low-risk, intermediate-risk, and high-risk groups were 4.9%, 20.4%, and 43.8%, respectively, showing a clear increasing risk gradient. After adding chronic kidney disease to the above trivariate scoring model, the ARROC only increased from 0.774 to 0.783, while the Brier score remained essentially unchanged, indicating that the predictive performance gain from adding additional variables was relatively limited. Therefore, the above method can achieve 30-day unplanned all-cause readmission risk stratification with less data input and simpler calculations.
[0064] The RCS analysis and Sullivan scoring described above are used to determine scoring parameters. These parameters do not need to be repeated when assessing risk in new patients; instead, pre-defined thresholds and scoring criteria can be directly applied. This approach maintains a certain level of risk identification capability and robustness even in independent external cohorts and different patient cohorts or data environments.
[0065] Example 2: Cross-institutional heterogeneous data application based on RDW-CV substitution decision
[0066] This embodiment illustrates the application process of the RDW-CV substitution judgment mechanism in the above method when the target medical institution does not routinely provide the RDW-SD field.
[0067] In the primary study cohort, RDW-SD and RDW-CV showed a significant positive correlation (Pearsonr = 0.69, P < 0.001). Further ROC analysis was performed with RDW-SD > 45 fL as the target state, and using the Youden exponent maximization principle, RDW-CV was determined to have a surrogate threshold of 13.45%.
[0068] In the independent external validation cohort, since the healthcare institutions in this cohort routinely report RDW-CV instead of RDW-SD, an RDW abnormality was determined by an RDW-CV > 13.45% and assigned a score of 1; otherwise, a score of 0 was assigned. The ALR score was then calculated in conjunction with active malignancy and length of hospital stay. A total of 2,554 patients were included in the external validation cohort. After risk assessment based on the aforementioned alternative thresholds, among the 2,554 independent external validation patients who did not routinely report RDW-SD, the ALR score AUROC was 0.677. The actual 30-day readmission rates for the low, intermediate, and high-risk groups were 7.6%, 17.0%, and 30.7%, respectively, still showing an increasing risk gradient. These results indicate that the above method can support a unified risk scoring workflow under different RDW reporting fields, providing a data adaptation basis for cross-institutional applications.
[0069] This embodiment demonstrates that when different medical institutions use different RDW reporting fields, by using preset field identification and substitution judgment, a unified ALR score and risk stratification can be completed without rebuilding the entire scoring model.
[0070] Example 3: Robustness verification in different patients and clinical settings
[0071] This embodiment is used to illustrate the robustness of the scoring method in different patient populations and clinical contexts.
[0072] The primary cohort included 2,254 participants in the non-tumor subgroup, while the external validation non-tumor subgroup comprised 2,102 participants. In this population, the active malignancy score was fixed at 0, with a corresponding ALR score range of 0–3, generated from length of hospital stay and RDW status. Considering the high weight of active malignancy in the ALR score, further validation was conducted in the non-tumor subgroup. Results showed that the AUROC of the ALR score in the primary cohort and the external cohort non-tumor subgroup were 0.678 and 0.657, respectively, and these scores were still able to distinguish between different risk levels.
[0073] Furthermore, after further adjustments for admission time and antiviral treatment during index hospitalization, the three core indicators—active malignancy, length of hospital stay, and RDW-SD—remained independently statistically associated with 30-day readmission:
[0074] The odds ratio (OR) for active malignant tumors was 9.66 (95% CI: 7.11–13.13).
[0075] The odds ratio (OR) was 1.02 for each additional day of hospital stay (95% CI: 1.02–1.03).
[0076] The odds ratio (OR) for each 1 fL increase in RDW-SD was 1.02 (95% CI: 1.00–1.03).
[0077] The above results provide supplementary experimental evidence for the robustness of the scoring scheme of the above method in different patient and epidemiological and treatment contexts.
Claims
1. A readmission risk assessment method based on heterogeneous electronic medical record data adaptation, wherein the method is executed by a computer system, characterized in that, Includes the following steps: Step S1, Data Acquisition and Feature Extraction: Obtain the target patient's medical record data during the indexed hospitalization and at discharge from the electronic medical record database, and extract the active malignant tumor status, hospitalization time, and red blood cell distribution width correlation test fields from the medical record data; Step S2: Field identification and adaptation, abnormal state determination: The red blood cell distribution width related test fields described in step S1 are searched and replaced: If the red blood cell distribution width standard deviation field is available, the value of the red blood cell distribution width standard deviation field is extracted and compared with 45fL. When the value of the red blood cell distribution width standard deviation field is greater than 45fL, the red blood cell distribution width state is determined to be an abnormal state; otherwise, the red blood cell distribution width state is determined to be a non-abnormal state. If the erythrocyte distribution width standard deviation field is missing or unavailable, the system automatically switches to the alternative decision path. If the erythrocyte distribution width coefficient of variation field is available, its value is extracted and compared to 13.45%. If the value of the erythrocyte distribution width coefficient of variation field is greater than 13.45%, the erythrocyte distribution width is determined to be in an abnormal state; otherwise, it is determined to be in a normal state. The 13.45% threshold is an alternative decision threshold pre-calculated based on the Youden index maximization principle, with the target state being a erythrocyte distribution width standard deviation field value greater than 45 fL. If both the standard deviation of red blood cell distribution width and the coefficient of variation of red blood cell distribution width are missing or unavailable, the red blood cell distribution width status is determined to be missing data. Step S3: Risk Score: Based on the active malignant tumor status and length of hospitalization described in Step S1, and the red blood cell distribution width status determined in Step S2, assign 5 points to the presence of active malignant tumor, 2 points to the length of hospitalization greater than 14 days, and 1 point to the abnormal red blood cell distribution width status. The above three scores are added together to generate a comprehensive re-admission risk score ranging from 0 to 8 points. Step S4, Risk Stratification Output: Based on the comprehensive re-admission risk score described in Step S3, the corresponding risk level is output according to the preset mapping relationship between the score range and the risk level. The risk level, the comprehensive re-admission risk score, the active malignant tumor status described in Step S1, the length of hospital stay, and the red blood cell distribution width status determined in Step S2 are output to the electronic medical record terminal, the discharge management system, or the follow-up management system.
2. The method according to claim 1, characterized in that, The active malignant tumor status mentioned in step S1 is identified based on the structured diagnostic fields or preset codes and field mappings in the electronic medical record database; the hospitalization time feature is used to extract the patient's admission time and discharge time, and calculate the actual hospitalization time in days; the red blood cell distribution width related test field is the available field and its corresponding value of the laboratory test field in the electronic medical record database.
3. The method according to claim 1, characterized in that, The mapping relationship between the preset scoring range and risk level mentioned in step S4 is as follows: a comprehensive score of 0 to 1 for readmission risk is classified as low risk, 2 to 5 as medium risk, and 6 to 8 as high risk.
4. A readmission risk assessment system based on heterogeneous electronic medical record data adaptation, characterized in that, include: The data acquisition and preprocessing module is used to acquire the target patient's medical record data during the indexed hospitalization and at discharge through the electronic medical record data interface, and extract the active malignant tumor status, hospitalization time and red blood cell distribution width related test fields from the medical record data; The field identification and adaptation module, communicatively connected to the data acquisition and preprocessing module, receives red blood cell distribution width (RBV) related test fields and performs retrieval and substitution determination on these fields. If the RBV standard deviation field is available, its value is extracted and compared with 45 fL. If the value is greater than 45 fL, the RBV status is determined to be abnormal; otherwise, it is determined to be normal. If the RBV standard deviation field is missing or unavailable, the module automatically switches to the substitution determination path to retrieve the RBV variation coefficient. If the data is available, the value of the red blood cell distribution width coefficient of variation field is extracted and compared with 13.45%. If the value of the red blood cell distribution width coefficient of variation field is greater than 13.45%, the red blood cell distribution width status is determined to be abnormal; otherwise, the red blood cell distribution width status is determined to be normal. The 13.45% is an alternative judgment threshold pre-calculated based on the principle of maximizing Youden's index, with the target status being a red blood cell distribution width standard deviation field value greater than 45 fL. If both the red blood cell distribution width standard deviation field and the red blood cell distribution width coefficient of variation field are missing or unavailable, the red blood cell distribution width status is determined to be missing data. The risk score calculation module is communicatively connected to the data acquisition and preprocessing module and the field identification and adaptation module, respectively. It is used to assign 5 points to the presence of active malignant tumor, 2 points to the state of hospitalization longer than 14 days, and 1 point to the state of abnormal red blood cell distribution width based on the status of active malignant tumor, length of hospitalization, and red blood cell distribution width. The above three scores are added together to generate a comprehensive re-admission risk score ranging from 0 to 8 points. The risk stratification output module is communicatively connected to the risk score calculation module. It is used to output the corresponding risk level according to the comprehensive re-admission risk score and the preset mapping relationship between the score range and the risk level. The risk level, comprehensive re-admission risk score, active malignant tumor status, length of hospital stay, and red blood cell distribution width status are output to the electronic medical record terminal, discharge management system, or follow-up management system.
5. A computer device comprising a processor and a memory, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the method described in any one of claims 1 to 3.
6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it causes the processor to implement the method described in any one of claims 1 to 3.