An apparatus for calculating morbidity based on medical record data and a storage medium
By acquiring and filtering medical record data and using outpatient case prediction models to analyze missing data, the problem of missing data in rare disease incidence statistics has been solved, and more accurate incidence rate calculations have been achieved.
Patent Information
- Application Number
- CN202610470435.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-04-10
- Publication Date
- 2026-07-03
AI Technical Summary
Existing methods for calculating the incidence of rare diseases fail to effectively utilize the differences between outpatient and inpatient records, resulting in inaccurate incidence estimates when medical data is lacking.
By acquiring medical records from multiple patients, a data filtering module was used to accurately identify patients with rare diseases. The number of missing outpatient cases was analyzed using an outpatient case prediction model. By utilizing the inherent correlation between outpatient and inpatient records, a regression model was established to calculate the incidence rate.
Even when medical data is missing, it can still achieve relatively reliable incidence rate estimates, thus improving the accuracy of rare disease incidence rates.
Smart Images
Figure CN122337664A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, and in particular to a device and storage medium for calculating the incidence rate based on medical record data. Background Technology
[0002] Rare diseases are a collective term for diseases with extremely low incidence rates. For example, a disease can be defined as a rare disease if it meets at least one of the following criteria: a neonatal incidence rate of less than 1 / 10,000, a prevalence rate of less than 1 / 10,000, and the number of patients less than 140,000. With the rapid accumulation of health and medical big data, the massive data resources from routine outpatient and inpatient medical records provide a new path for the intelligent estimation of rare disease incidence rates.
[0003] However, in practical applications, existing rare disease incidence statistics do not consider the differences between outpatient and inpatient records at different stages of disease progression, thus neglecting the value of data from different types of visits, such as outpatient and inpatient data. Therefore, if medical data is incomplete, such as some medical institutions lacking outpatient data or having incomplete data coverage, the final estimated incidence rate will be inaccurate. Summary of the Invention
[0004] This application provides a device and storage medium for calculating the incidence rate based on medical record data, so as to improve the accuracy of the final obtained incidence rate of rare diseases.
[0005] In a first aspect, embodiments of this application provide a device for calculating incidence rates based on medical record data, comprising: The data acquisition module is used to acquire medical records data from multiple patients; The data filtering module is used to filter the medical record data of the multiple patients to obtain the medical record data of the target patient suffering from a rare disease. The medical record data of the target patient includes the identifier of the medical institution where the target patient received treatment. The medical institution includes a first category of medical institutions and a second category of medical institutions. The first category of medical institutions is used to provide the number of first outpatient cases and the number of first inpatient cases suffering from the rare disease, and the second category of medical institutions is used to provide the number of second inpatient cases suffering from the rare disease. The data prediction module is used to analyze the second number of inpatient cases based on the first number of outpatient cases and the first number of inpatient cases using an outpatient case prediction model, and obtain the prediction result of the second number of outpatient cases with the rare disease in the second category of medical institutions. The outpatient case prediction model is used to describe the correlation between the number of outpatient cases and the number of inpatient cases with the rare disease in the same medical institution. The first data determination module is used to determine the incidence rate of the rare disease based on the medical record data of the target patient and the prediction results.
[0006] Optionally, the data prediction module is further configured to: Obtain the structural characteristics of the first category of medical institutions, which are used to describe the size, level, regional attributes and service capabilities of the first category of medical institutions. A regression model was established using the number of the first inpatient cases as the independent variable, the number of the first outpatient cases as the dependent variable, and the structural characteristics of the medical institutions in the first category as covariates. The regression model is used as the outpatient case prediction model.
[0007] Optionally, the data prediction module is specifically used for: Determine the variance and mean of the first number of outpatient cases; When the difference between the variance and the mean is less than the difference threshold, a Poisson regression model is adopted, with the number of the first inpatient cases as the independent variable, the number of the first outpatient cases as the dependent variable, and the structural characteristics of the medical institutions of the first category as the covariates, to establish the regression model. If the difference between the variance and the mean is greater than the difference threshold, a negative binomial regression model is adopted, with the number of the first inpatient cases as the independent variable, the number of the first outpatient cases as the dependent variable, and the structural characteristics of the medical institutions of the first category as covariates, to establish the regression model.
[0008] Optionally, the data filtering module is specifically used for: The medical records of the multiple patients are identified based on a first disease identification rule to obtain identification results for each of the multiple patients; the first disease identification rule includes multiple disease classification codes and identification restrictions corresponding to each of the multiple disease classification codes. For the first patient among the plurality of patients, if the identification result corresponding to the first patient indicates that the first patient's medical record data matches the first disease classification code among the plurality of disease classification codes and satisfies the identification restriction conditions of the first disease classification code, the medical record data of the first patient is obtained from the medical record data of the plurality of patients and used as the medical record data of the target patient.
[0009] Optionally, the device for calculating the incidence rate based on medical record data further includes: The feature extraction module is used to extract features from the multidimensional medical data of the target patient to obtain multiple medical features corresponding to the multidimensional medical data. The multidimensional medical data includes, in addition to medical record data, the target patient's disease course data, clinical behavior data, medication data, or medical test data. The rule generation module is used to generate a second disease identification rule corresponding to each disease classification code based on each of the plurality of disease classification codes and at least one of the plurality of medical features. The second disease identification rule corresponding to each disease classification code includes identification restrictions corresponding to each disease classification code and the at least one medical feature. The data recognition module is used to recognize the medical record data of the target patient based on the second disease recognition rule corresponding to each disease classification code, to obtain a first case recognition result, and to obtain a second case recognition result corresponding to the medical record data of the target patient, wherein the second case recognition result is manually labeled. The performance determination module is used to determine the performance index of the second disease identification rule based on the first case identification result and the second case identification result. The performance index is used to indicate the identification performance of the second disease identification rule.
[0010] Optionally, the data filtering module is specifically used for: When the performance index is greater than or equal to the performance index threshold, the medical record data of the multiple patients are identified based on the second disease identification rule to obtain the identification results corresponding to the multiple patients respectively; If the identification result corresponding to the first patient indicates that the first patient's medical record data matches the first disease classification code among the multiple disease classification codes and satisfies the identification restriction conditions of the first disease classification code and the identification restriction conditions in the second disease identification rules, the first patient's medical record data is obtained from the medical record data of the multiple patients and used as the medical record data of the target patient.
[0011] Optionally, the device for calculating the incidence rate based on medical record data further includes: The second data determination module is used to determine the first medical record data of the same patient based on the visit time in the multiple medical record data of the same patient when the medical record data of the target patient includes multiple medical record data of the same patient. The first data determination module is specifically used for: The incidence rate of the rare disease is determined based on the initial medical record data of the same patient and the prediction results.
[0012] Optionally, the data acquisition module is specifically used for: Obtain the initial medical records of the aforementioned patients; The initial medical record data of the multiple patients are preprocessed to obtain the intermediate medical record data of the multiple patients. The identity information in the intermediate medical record data of the multiple patients is de-identified and encoded to obtain the medical record data of the multiple patients, and the medical record data of the multiple patients includes the privacy identifiers corresponding to the identity information of the multiple patients.
[0013] Secondly, embodiments of this application provide an electronic device, the device including: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store a program, the program including instructions that, when executed by the processor, cause the processor to perform any of the implementation steps performed by the modules included in the device for calculating the incidence rate based on medical record data.
[0014] Thirdly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any of the implementation steps performed by the modules included in the device for calculating the incidence rate based on medical record data.
[0015] As can be seen from the above technical solutions, the embodiments of this application have the following advantages: In this embodiment, a device for calculating the incidence rate based on medical record data can be provided, comprising a data acquisition module, a data filtering module, a data prediction module, and a first data determination module. The data acquisition module acquires medical record data from multiple patients; the data filtering module filters the medical record data from multiple patients to obtain the medical record data of a target patient suffering from a rare disease. The target patient's medical record data includes the identifier of the medical institution where the target patient sought treatment. These medical institutions are categorized into first-class and second-class medical institutions. The first-class medical institutions provide the number of first outpatient cases and first inpatient cases of rare diseases, while the second-class medical institutions provide the number of second inpatient cases of rare diseases. The data prediction module analyzes the number of second inpatient cases based on the first outpatient and first inpatient cases using an outpatient case prediction model to obtain a prediction result for the number of second outpatient cases of rare diseases at the second-class medical institutions. The outpatient case prediction model describes the correlation between the number of outpatient and inpatient cases of rare diseases within the same medical institution. The first data determination module determines the incidence rate of the rare disease based on the target patient's medical record data and the prediction result.
[0016] As can be seen, by collecting medical records from multiple patients through the data acquisition module, and then relying on the data filtering module, the medical records of target patients with rare diseases can be accurately filtered out from these records. Furthermore, due to differences in data coverage among different medical institutions, Category I medical institutions can provide the number of outpatient and inpatient cases corresponding to rare diseases (i.e., the number of first outpatient cases and the number of first inpatient cases), while Category II medical institutions can only provide the number of inpatient cases (i.e., the number of second inpatient cases) and lack the corresponding number of second outpatient cases. Therefore, for the situation where Category II medical institutions lack outpatient cases, the data prediction module and the complete case count provided by Category I medical institutions can be used to analyze the known number of second inpatient cases using an outpatient case prediction model. This allows for the accurate prediction of the number of second outpatient cases by fully utilizing the value of single inpatient data, based on the differences and inherent correlations between outpatient and inpatient records at different stages of disease progression. Finally, the first data determination module determines the incidence rate of the rare disease based on the target patient's medical records and the predicted number of second outpatient cases. In this way, even when medical data is missing, a relatively reliable incidence rate estimation scheme can still be achieved, thereby improving the accuracy of the final incidence rate of rare diseases. Attached Figure Description
[0017] Figure 1 A schematic diagram of the structure of a device for calculating the incidence rate based on medical record data, provided in an embodiment of this application; Figure 2 A schematic diagram of another device for calculating incidence rates based on medical record data provided in this application embodiment; Figure 3 This is a schematic diagram of the structure of another device for calculating the incidence rate based on medical record data, provided in an embodiment of this application. Detailed Implementation
[0018] As mentioned earlier, existing rare disease incidence statistics do not consider the differences between outpatient and inpatient records at different stages of disease progression, thus ignoring the value of data from different types of visits, such as outpatient and inpatient data. Therefore, if medical data is incomplete, such as some medical institutions lacking outpatient data or having incomplete data coverage, the final estimated incidence rate will be inaccurate.
[0019] To address the aforementioned issues, this application provides a device for calculating incidence rates based on medical record data, comprising a data acquisition module, a data filtering module, a data prediction module, and a first data determination module. The data acquisition module acquires medical record data from multiple patients. The data filtering module filters the medical record data from multiple patients to obtain the medical record data of a target patient suffering from a rare disease. The target patient's medical record data includes the identifier of the medical institution where the target patient sought treatment. These medical institutions are categorized into first-class and second-class institutions. The first-class institutions provide the number of first outpatient and first inpatient cases of rare diseases, while the second-class institutions provide the number of second inpatient cases of rare diseases. The data prediction module analyzes the number of second inpatient cases based on the first outpatient and first inpatient cases using an outpatient case prediction model to obtain a prediction result for the number of second outpatient cases of rare diseases at the second-class medical institutions. The outpatient case prediction model describes the correlation between the number of outpatient and inpatient cases of rare diseases within the same medical institution. The first data determination module determines the incidence rate of the rare disease based on the target patient's medical record data and the prediction result.
[0020] As can be seen, by collecting medical records from multiple patients through the data acquisition module, and then relying on the data filtering module, the medical records of target patients with rare diseases can be accurately filtered out from these records. Furthermore, due to differences in data coverage among different medical institutions, Category I medical institutions can provide the number of outpatient and inpatient cases corresponding to rare diseases (i.e., the number of first outpatient cases and the number of first inpatient cases), while Category II medical institutions can only provide the number of inpatient cases (i.e., the number of second inpatient cases) and lack the corresponding number of second outpatient cases. Therefore, for the situation where Category II medical institutions lack outpatient cases, the data prediction module and the complete case count provided by Category I medical institutions can be used to analyze the known number of second inpatient cases using an outpatient case prediction model. This allows for the accurate prediction of the number of second outpatient cases by fully utilizing the value of single inpatient data, based on the differences and inherent correlations between outpatient and inpatient records at different stages of disease progression. Finally, the first data determination module determines the incidence rate of the rare disease based on the target patient's medical records and the predicted number of second outpatient cases. In this way, even when medical data is missing, a relatively reliable incidence rate estimation scheme can still be achieved, thereby improving the accuracy of the final incidence rate of rare diseases.
[0021] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0022] Figure 1 This is a schematic diagram of a device for calculating incidence rates based on medical record data, provided as an embodiment of this application. (Combined with...) Figure 1 As shown in the embodiments of this application, the device for calculating incidence rates based on medical record data may include a data acquisition module, a data filtering module, a data prediction module, and a first data determination module. For ease of understanding, the function of each module, i.e., the steps performed by each module, will be described exemplarily below.
[0023] Regarding the data acquisition module, this module can be used to acquire medical record data from multiple patients. Medical record data refers to health-related data generated when a patient visits a medical institution, such as outpatient records, admission records, or discharge records. Each data entry can include fields such as the patient's original identity information (e.g., gender, age, and date of birth), disease classification code, date of visit, type of visit, and identification of the medical institution visited. Here, the disease classification code, for example, is ICD-10 coding (a standardized disease coding system), which can be used for standardized disease diagnosis and statistics.
[0024] Therefore, in specific implementation, the data acquisition module may include a data interface, a data preprocessing module, and a privacy coding module.
[0025] Accordingly, for the implementation of obtaining medical record data of multiple patients, firstly, the data acquisition module can connect to the information systems of multiple medical institutions through a data interface to obtain the initial medical record data of multiple patients.
[0026] Next, the initial medical record data is preprocessed by the data preprocessing module to obtain intermediate medical record data for multiple patients. Preprocessing can refer to operations such as format unification, field standardization, duplicate record merging, outlier removal, and missing value marking on data from different sources and with different structures, in order to obtain standardized intermediate medical record data with a uniform structure and computability.
[0027] It should be noted that the implementation process of the above preprocessing operations is not specifically limited in the embodiments of this application, and can be implemented by any existing or future preprocessing method.
[0028] Then, the privacy coding module is used to de-identify and encode the identity information in the intermediate medical record data of multiple patients, resulting in medical record data for multiple patients. De-identification coding refers to processing the original identity information through encryption or hashing to generate a privacy identifier for cross-record association. This identifier can achieve patient record matching across medical institutions and data sources without revealing the patient's true identity, thus obtaining the privacy identifier corresponding to each patient's identity information. In other words, the medical record data of the aforementioned multiple patients can include the privacy identifier corresponding to each patient's identity information.
[0029] In this way, the data acquisition module, through the cooperation of the data interface, data preprocessing module, and privacy coding module, can provide a structured and computable data foundation for subsequent data screening and incidence rate calculation while protecting patient privacy, thus avoiding the disruption of the calculation process due to chaotic data formats or privacy leaks.
[0030] Regarding the data filtering module, this module can be used to filter the medical record data of multiple patients to obtain the medical record data of target patients suffering from rare diseases. Here, the target patient refers to a patient with a rare disease among multiple patients. Accordingly, various possible implementation methods can be provided for this data filtering module in the embodiments of this application, which are described below by example.
[0031] As one possible implementation, when the data filtering module performs filtering, it can first identify the medical records of multiple patients based on a first disease identification rule, obtaining identification results for each patient. The first disease identification rule can include multiple disease classification codes mentioned above and identification restrictions corresponding to each disease classification code.
[0032] This is because, for rare diseases with clearly defined ICD-10 codes that uniquely correspond to a specific rare disease, accurate identification can be achieved directly based on the diagnostic code without additional identification restrictions. However, for rare diseases with a broad ICD-10 code range that cannot be used as the sole basis for judgment, secondary confirmation is required based on code matching and additional identification restrictions to improve the specificity and accuracy of case identification. For ease of understanding, the first disease identification rule is illustrated below with reference to Table 1.
[0033] Table 1
[0034] As shown in Table 1, the corresponding disease classification code for Angelman's syndrome is Q93.501, with no additional identification restrictions. Therefore, as long as the medical record data includes the above code, the patient can be identified as a target patient with Angelman's syndrome. For hyperhomocysteinemia, the corresponding disease classification codes are E72.101 and E72.100x004, and they have additional identification restrictions: the patient must simultaneously meet two or more diagnoses with an interval of 90 days or more, and be under 18 years of age. In other words, the medical record data must include the above code and meet the above conditions for the patient to be identified as a target patient with hyperhomocysteinemia.
[0035] Therefore, taking the first patient among multiple patients as an example, if the identification result corresponding to the first patient indicates that the first patient's medical record data matches the first disease classification code among multiple disease classification codes and meets the identification restriction conditions of the first disease classification code, the data filtering module can obtain the first patient's medical record data from the medical record data of multiple patients and use it as the medical record data of the target patient.
[0036] As can be seen, by using the above screening methods, relying on the first disease classification code confirmed by clinical experts and additional identification restrictions, it is possible to accurately extract data on patients with rare diseases, avoid the inclusion of non-target cases, and improve the accuracy of subsequent statistics.
[0037] As another possible implementation, since different medical institutions have certain differences in coding standards, treatment pathways and data integrity, relying solely on diagnostic coding may easily lead to misidentification and omission. Therefore, in order to further improve the accuracy, stability and generalization ability of rare disease case screening, in this embodiment of the application, a computable phenotypic model that integrates multidimensional information can be constructed, and a gold standard-driven performance evaluation mechanism for disease identification rules can be introduced to obtain a second case identification rule that can improve the quality of case identification, thereby improving the reliability and scientificity of the incidence rate estimation results.
[0038] For ease of understanding, the performance evaluation mechanism will first be described by way of example in the embodiments of this application.
[0039] Figure 2 A schematic diagram of another device for calculating incidence rates based on medical record data, provided as an embodiment of this application. (Combined with...) Figure 2 As shown, in specific implementation, the device for calculating the incidence rate based on medical record data may also include: a feature extraction module, a rule generation module, a data recognition module, and a performance determination module.
[0040] Accordingly, firstly, the feature extraction module extracts features from the target patient's multidimensional medical data, obtaining multiple medical features corresponding to the multidimensional medical data. This multidimensional medical data can include, in addition to medical records, the target patient's disease progression data, clinical behavior data, medication data, or medical test data. In this way, by incorporating data related to disease progression, medication, laboratory tests, and clinical behavior, the target patient's medical characteristics can be characterized in multiple dimensions, facilitating the formation of a structured and computable set of phenotypic features.
[0041] Next, the rule generation module is used to generate a second disease identification rule corresponding to each disease classification code, based on each disease classification code and at least one medical feature from multiple medical features. The second disease identification rule for each disease classification code includes the corresponding disease classification code and identification constraints corresponding to at least one medical feature. It is evident that the second disease identification rule can add corresponding identification constraints based on medical features, building upon the disease classification code, thereby integrating computable phenotypic features of multidimensional information into the disease identification rule, which helps to achieve more accurate identification.
[0042] For example, assuming there are medical features a, b, and c, and the disease classification code corresponding to a rare disease is x1.y1, then the second disease identification rule corresponding to this disease classification code may include at least one of the following: Second disease identification rule 1 may include the disease classification code x1.y1 and medical feature a; Second disease identification rule 2 may include the disease classification code x1.y1 and medical feature a+b; Second disease identification rule 3 may include the disease classification code x1.y1 and medical feature a+b+c; Second disease identification rule 4 may include the disease classification code x1.y1 and medical feature b+c.
[0043] Then, through the data recognition module, the medical record data of the target patient can be identified based on the second disease recognition rule to obtain the first case recognition result. Here, the first case recognition result refers to the result obtained based on at least one of the aforementioned second disease recognition rules. Therefore, each first case recognition result can reflect the performance of the corresponding second disease recognition rule.
[0044] Furthermore, the data identification module can further obtain the second case identification results corresponding to the target patient's medical record data. Here, the second case identification results are manually labeled. That is to say, in practical applications, the second case identification results can be reviewed by clinical experts or epidemiologists on the target patient's medical record data, combining diagnostic records, disease descriptions, medication information, test results, and clinical behavior to comprehensively determine whether it is a genuine rare disease case. For cases with discrepancies, final confirmation can be achieved through expert review or multi-expert consultation, forming a highly reliable labeled dataset that has been reviewed by experts. Therefore, the second case identification results can serve as the gold standard data for performance evaluation of second disease identification rules.
[0045] Finally, using the performance determination module, the performance metrics of the second disease identification rule are determined based on the results of the first and second case identifications. These performance metrics are indicators that demonstrate the identification performance of the second disease identification rule. For example, performance metrics for the second disease identification rule may include sensitivity (SEN) and positive predictive value (PPV). In other words, in practical applications, the gold standard described above can be used as a reference to compare the results of at least one case identification, thereby clarifying the identification performance of at least one of the above second disease identification rules.
[0046] For ease of understanding, the following example uses the second disease identification rule 1 corresponding to a disease classification code, and is illustrated in Table 2.
[0047] Table 2
[0048] As shown in Table 2 above, the number of cases where both the first and second case identification results indicate that the disease is a rare disease is denoted as A; the number of cases where the first case identification result indicates that the disease is a rare disease but the second case identification result indicates that the disease is not a rare disease is denoted as B; the number of cases where the first case identification result indicates that the disease is not a rare disease but the second case identification result indicates that the disease is a rare disease is denoted as C; and the number of cases where both the first and second case identification results indicate that the disease is not a rare disease is denoted as D.
[0049] Additionally, it should be noted that the first case identification results shown in Table 2 above are based on the results determined by the second disease identification rule 1. That is, the identification performance of the second disease identification rule 1 can be obtained by calculating the performance indicators.
[0050] Accordingly, the SEN of the second disease identification rule 1 can be calculated using the following formula (1): SEN=A / (A+C) (1) The PPV of the second disease identification rule 1 can be calculated using the following formula (2): PPV = A / (A+B) (2) In this way, after evaluating the performance of the second disease identification rule using the above performance evaluation mechanism, the data filtering module can further select a suitable second disease identification rule for data identification based on the performance evaluation results.
[0051] Specifically, the data filtering module can identify the medical records of multiple patients based on the second disease identification rule when the performance index is greater than or equal to the performance index threshold, and obtain the identification results corresponding to each patient.
[0052] Taking the first patient as an example, if the identification result corresponding to the first patient indicates that the first patient's medical record data matches the first disease classification code among multiple disease classification codes and satisfies the identification restriction conditions of the first disease classification code and the identification restriction conditions in the second disease identification rules, then the data filtering module can obtain the first patient's medical record data from the medical record data of multiple patients and use it as the medical record data of the target patient.
[0053] Thus, the disease identification process requires not only that the first patient's medical record data meet the first disease classification code and its corresponding identification constraints, but also that it meets the identification constraints in the second disease identification rules. Therefore, this reduces the risk of misidentification and missed identification caused by single-code identification, further improving the accuracy of case screening.
[0054] As another possible implementation, the data filtering module described above can also use a machine learning model trained on historical labeled data to automatically identify and obtain the medical records of the target patient. Here, historical labeled data refers to historically identified rare disease case data.
[0055] Alternatively, the data filtering module can employ a hybrid recognition architecture that combines third disease classification rules with machine learning to comprehensively assess patients. Here, the third disease classification rules are similar to the first or second disease classification rules mentioned above, and may include ICD-10 codes and recognition restrictions such as age limits, frequency of visits, and interval requirements.
[0056] Accordingly, the medical record data can be initially screened using the third disease classification rules to obtain intermediate data; then, the intermediate data is input into a trained machine learning model to identify the medical record data of the target patient. It is evident that this hybrid identification architecture can balance the interpretability of the third disease classification rules with the generalization ability of machine learning, thereby improving identification stability while enhancing the accuracy of identifying complex cases.
[0057] It should be noted that, for the above-mentioned data filtering modules, any of the above methods can be selected for implementation based on actual needs or business scenarios, so that the medical record data of the target patients can be identified based on the same implementation method.
[0058] As mentioned earlier, medical record data can include the identifier of the medical institution the patient visited. Therefore, the medical record data of the target patient filtered by the data filtering module described above can include the identifier of the medical institution visited by that target patient.
[0059] In practical applications, the medical institutions that patients currently visit can be divided into two categories: one category has complete outpatient and inpatient data, and can provide the number of first outpatient cases and first inpatient cases of rare diseases, which can be referred to as the first category of medical institutions; the other category only has inpatient data and lacks complete outpatient data, that is, it only provides the number of second inpatient cases of rare diseases, which can be referred to as the second category of medical institutions.
[0060] Based on this, under the aforementioned real-world conditions, directly calculating the incidence rate based on existing data would obviously lead to a systematic underestimation and structural bias, making the calculation of the incidence rate unreliable. Therefore, in this embodiment of the application, the data prediction module can be used to reasonably infer and complete the data of medical institutions with missing data (i.e., the medical institutions of the second category mentioned above), thereby reducing the structural underestimation problem caused by missing data.
[0061] In practical implementation, the data prediction module can be used to analyze the number of second inpatient cases based on the number of first outpatient cases and the number of first inpatient cases through an outpatient case prediction model, thereby obtaining the predicted number of second outpatient cases with rare diseases in the second category of medical institutions.
[0062] The outpatient case prediction model refers to a model capable of predicting and completing missing outpatient data for medical institutions in the second category. Therefore, in this embodiment, the outpatient case prediction model can be used to describe the correlation between the number of outpatient cases and inpatient cases of rare diseases in the same medical institution. For ease of understanding, the construction process of the outpatient case prediction model will be described below as an example.
[0063] In practical implementation, the data prediction module can also first obtain the structural characteristics of the first category of medical institutions. Understandably, structural characteristics are fixed features used to objectively describe the attributes and service capabilities of a medical institution, and do not change with short-term fluctuations in patient visit data. Therefore, in this embodiment, the aforementioned structural characteristics of the first category of medical institutions can be used to describe the scale, level, regional attributes, and service capabilities of the first category of medical institutions.
[0064] Among these, the scale of a medical institution refers to indicators reflecting its size, such as the number of beds, annual outpatient volume, and the population it serves; the grade of a medical institution refers to the hospital level approved by the health authorities, reflecting its diagnostic and treatment capabilities and resource allocation; the regional attributes of a medical institution refer to its positioning within the regional healthcare system, including whether it is a regional medical center, whether it leads or participates in the construction of medical consortia, and the distribution of medical resources in its region; and the service capacity of a medical institution refers to its scope of diagnostic and treatment services, its ability to access payment methods, and its level of information technology services, specifically including whether it has implemented reforms to the diagnosis-related group (DRG) payment method, whether it has implemented reforms to the medical insurance payment method based on disease type, whether it has achieved direct settlement with medical insurance agencies, and whether it provides internet-based medical services.
[0065] After obtaining the structural characteristics of the first category of medical institutions, the data prediction module then uses the number of the first inpatient cases as the independent variable, the number of the first outpatient cases as the dependent variable, and the structural characteristics of the first category of medical institutions as the covariates to establish a regression model, and uses the regression model as the outpatient case prediction model.
[0066] When establishing the model, firstly, determine the variance and mean of the number of outpatient cases. If the difference between the variance and the mean is less than the difference threshold, it indicates that the data dispersion is low. Therefore, a Poisson regression model can be used to construct the regression model. If the difference between the variance and the mean is greater than the difference threshold, it indicates that the data has excessive dispersion. Therefore, a negative binomial regression model can be used to construct the regression model. The above two models can stably depict the correlation between the number of outpatient cases and the number of inpatient cases, which helps to establish a model with better predictive performance. For ease of understanding, the outpatient case prediction model will be introduced below in conjunction with formula (3).
[0067] (3)
[0068] in, Indicates medical institutions In the year The average number of outpatient cases, This indicates the number of inpatient cases at the medical institution in the corresponding year. Its corresponding regression coefficient, For the intercept term, The regression coefficients are the structural characteristics of the medical institution.
[0069] In this way, after the model training is completed, the number of second inpatient cases and the structural characteristics of the medical institutions of the second institution can be input into the above model to obtain the prediction results of the number of second outpatient cases.
[0070] It is understandable that the above outpatient case prediction model is modeled using a regression model as an example. In practical applications, Bayesian inference models, hierarchical models, or other statistical learning models can also be used, and no specific limitation is required here.
[0071] Furthermore, regarding the first data determination module, this module can be used to determine the incidence rate of rare diseases based on the target patient's medical record data and prediction results. The incidence rate of a rare disease refers to the frequency of new cases of a particular rare disease occurring in a certain population within a certain period.
[0072] Accordingly, when calculating the incidence rate in the first data determination module, the number of first outpatient cases and first inpatient cases directly observed by first-category medical institutions can be combined with the predicted results of the number of second inpatient cases and second outpatient cases from second-category medical institutions to obtain the total number of new rare disease cases in the region. The total number of new cases is then divided by the corresponding population size to complete the incidence rate calculation. In this way, the incidence rate calculation process integrates measured and predicted data, overcoming the limitations of incomplete data. Therefore, it can output accurate incidence rate results, providing data support for rare disease epidemiological analysis and policy formulation.
[0073] Furthermore, in practical applications, since the same patient may visit different medical institutions, and different types of visits (such as outpatient or inpatient) may generate multiple records, directly performing record-level statistics would inevitably lead to duplicate counting of the number of cases, thereby overestimating the disease burden. Therefore, the embodiments of this application can also determine the first onset event node at the individual level through the first-visit determination mechanism, changing the statistical unit from "medical records" to "individual events," thereby improving the accuracy of the statistical results of the number of cases.
[0074] Based on this Figure 3 This is a schematic diagram of another device for calculating incidence rates based on medical record data, provided as an embodiment of this application. (Combined with...) Figure 3 As shown, in specific implementation, the device for calculating the incidence rate based on medical record data may further include a second data determination module. This second data determination module can be used to determine the first medical record data of the same patient based on the visit time in the multiple medical record data of the same patient when the target patient's medical record data includes multiple medical record data of the same patient. The first medical record data refers to the first visit event in which the patient is identified as having a rare disease by the medical system. By using the first medical record data, the onset time point can be defined, thereby avoiding duplicate counting.
[0075] Accordingly, the aforementioned first data determination module can determine the incidence rate of rare diseases based on the initial medical record data and prediction results of the same patient. For the method of determining the incidence rate, please refer to the relevant description in the above embodiments, which will not be elaborated here. In this way, by using each patient's first onset event as the statistical unit, replacing traditional record-level statistics, the problem of duplicate counting caused by multiple medical visits can be reduced, thereby significantly improving the accuracy of incidence rate calculation and solving the deficiency of large errors in the determination of onset events in existing technologies.
[0076] It should be noted that, since the aforementioned medical record data involves the privacy data of relevant users, when the embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions.
[0077] Based on the above content regarding the device for calculating incidence rates based on medical record data, it is clear that in this embodiment, the data acquisition module collects medical record data from multiple patients, and the data filtering module can then accurately filter out the medical record data of target patients suffering from rare diseases from this data. Furthermore, due to differences in data coverage among different medical institutions, first-category medical institutions can provide the number of outpatient and inpatient cases corresponding to rare diseases (i.e., the number of first outpatient cases and the number of first inpatient cases), while second-category medical institutions can only provide the number of inpatient cases corresponding to rare diseases (i.e., the number of second inpatient cases) and lack the corresponding number of second outpatient cases. Therefore, for the situation where the number of outpatient cases is missing in second-category medical institutions, the data prediction module and the complete number of cases provided by first-category medical institutions can be used to analyze the known number of second inpatient cases using an outpatient case prediction model. This allows for the accurate prediction of the number of second outpatient cases by fully utilizing the value of single inpatient data, based on the differences and inherent correlations between outpatient and inpatient records at the disease progression stage. Finally, the first data determination module determines the incidence rate of the rare disease based on the medical record data of the target patient and the prediction results of the number of second outpatient cases. In this way, even when medical data is missing, a relatively reliable incidence rate estimation scheme can still be achieved, thereby improving the accuracy of the final incidence rate of rare diseases.
[0078] Furthermore, embodiments of this application also provide an electronic device, including: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store one or more programs, the one or more programs including instructions that, when executed by the processor, cause the processor to perform any of the implementation steps performed by the modules included in the device for calculating the incidence rate based on medical record data.
[0079] Furthermore, embodiments of this application also provide a computer-readable storage medium storing instructions that, when executed on an electronic device, cause any implementation step to be performed by the modules included in the device for calculating the incidence rate based on medical record data.
[0080] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that all or part of the steps in the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network communication device such as a media gateway, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application. It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on describing the differences from other embodiments. The same or similar parts between the various embodiments can be referred to mutually.
[0081] The system disclosed in the embodiments is described in a relatively simple manner because it corresponds to the method disclosed in the embodiments. For relevant details, please refer to the method section.
[0082] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0083] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A device for calculating incidence rates based on medical record data, characterized in that, include: The data acquisition module is used to acquire medical records data from multiple patients; The data filtering module is used to filter the medical record data of the multiple patients to obtain the medical record data of the target patient suffering from a rare disease. The medical record data of the target patient includes the identifier of the medical institution where the target patient received treatment. The medical institution includes a first category of medical institutions and a second category of medical institutions. The first category of medical institutions is used to provide the number of first outpatient cases and the number of first inpatient cases suffering from the rare disease, and the second category of medical institutions is used to provide the number of second inpatient cases suffering from the rare disease. The data prediction module is used to analyze the second number of inpatient cases based on the first number of outpatient cases and the first number of inpatient cases using an outpatient case prediction model, and obtain the prediction result of the second number of outpatient cases with the rare disease in the second category of medical institutions. The outpatient case prediction model is used to describe the correlation between the number of outpatient cases and the number of inpatient cases with the rare disease in the same medical institution. The first data determination module is used to determine the incidence rate of the rare disease based on the medical record data of the target patient and the prediction results.
2. The device for calculating incidence rates based on medical record data according to claim 1, characterized in that, The data prediction module is also used for: Obtain the structural characteristics of the first category of medical institutions, which are used to describe the size, level, regional attributes and service capabilities of the first category of medical institutions. A regression model was established using the number of the first inpatient cases as the independent variable, the number of the first outpatient cases as the dependent variable, and the structural characteristics of the medical institutions in the first category as covariates. The regression model is used as the outpatient case prediction model.
3. The device for calculating incidence rates based on medical record data according to claim 2, characterized in that, The data prediction module is specifically used for: Determine the variance and mean of the first number of outpatient cases; When the difference between the variance and the mean is less than the difference threshold, a Poisson regression model is adopted, with the number of the first inpatient cases as the independent variable, the number of the first outpatient cases as the dependent variable, and the structural characteristics of the medical institutions of the first category as the covariates, to establish the regression model. If the difference between the variance and the mean is greater than the difference threshold, a negative binomial regression model is adopted, with the number of the first inpatient cases as the independent variable, the number of the first outpatient cases as the dependent variable, and the structural characteristics of the medical institutions of the first category as covariates, to establish the regression model.
4. The device for calculating incidence rates based on medical record data according to claim 1, characterized in that, The data filtering module is specifically used for: The medical records of the multiple patients are identified based on a first disease identification rule to obtain identification results for each of the multiple patients; the first disease identification rule includes multiple disease classification codes and identification restrictions corresponding to each of the multiple disease classification codes. For the first patient among the plurality of patients, if the identification result corresponding to the first patient indicates that the first patient's medical record data matches the first disease classification code among the plurality of disease classification codes and satisfies the identification restriction conditions of the first disease classification code, the medical record data of the first patient is obtained from the medical record data of the plurality of patients and used as the medical record data of the target patient.
5. The device for calculating incidence rates based on medical record data according to claim 4, characterized in that, The device for calculating the incidence rate based on medical record data also includes: The feature extraction module is used to extract features from the multidimensional medical data of the target patient to obtain multiple medical features corresponding to the multidimensional medical data. The multidimensional medical data includes, in addition to medical record data, the target patient's disease course data, clinical behavior data, medication data, or medical test data. The rule generation module is used to generate a second disease identification rule corresponding to each disease classification code based on each of the plurality of disease classification codes and at least one of the plurality of medical features. The second disease identification rule corresponding to each disease classification code includes identification restrictions corresponding to each disease classification code and the at least one medical feature. The data recognition module is used to recognize the medical record data of the target patient based on the second disease recognition rule corresponding to each disease classification code, to obtain a first case recognition result, and to obtain a second case recognition result corresponding to the medical record data of the target patient, wherein the second case recognition result is manually labeled. The performance determination module is used to determine the performance index of the second disease identification rule based on the first case identification result and the second case identification result. The performance index is used to indicate the identification performance of the second disease identification rule.
6. The device for calculating incidence rates based on medical record data according to claim 5, characterized in that, The data filtering module is specifically used for: When the performance index is greater than or equal to the performance index threshold, the medical record data of the multiple patients are identified based on the second disease identification rule to obtain the identification results corresponding to the multiple patients respectively; If the identification result corresponding to the first patient indicates that the first patient's medical record data matches the first disease classification code among the multiple disease classification codes and satisfies the identification restriction conditions of the first disease classification code and the identification restriction conditions in the second disease identification rules, the first patient's medical record data is obtained from the medical record data of the multiple patients and used as the medical record data of the target patient.
7. The device for calculating incidence rates based on medical record data according to any one of claims 1 to 6, characterized in that, The device for calculating the incidence rate based on medical record data also includes: The second data determination module is used to determine the first medical record data of the same patient based on the visit time in the multiple medical record data of the same patient when the medical record data of the target patient includes multiple medical record data of the same patient. The first data determination module is specifically used for: The incidence rate of the rare disease is determined based on the initial medical record data of the same patient and the prediction results.
8. The device for calculating incidence rates based on medical record data according to any one of claims 1 to 6, characterized in that, The data acquisition module is specifically used for: Obtain the initial medical records of the aforementioned patients; The initial medical record data of the multiple patients are preprocessed to obtain the intermediate medical record data of the multiple patients. The identity information in the intermediate medical record data of the multiple patients is de-identified and encoded to obtain the medical record data of the multiple patients, and the medical record data of the multiple patients includes the privacy identifiers corresponding to the identity information of the multiple patients.
9. A device for calculating incidence rates based on medical record data, characterized in that, The device includes: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store a program, the program including instructions that, when executed by the processor, cause the processor to perform the steps performed by the modules included in the device for calculating incidence rates based on medical record data according to any one of claims 1 to 8.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps performed by the modules included in the device for calculating incidence rates based on medical record data as described in any one of claims 1 to 8.