Multi-center clinical research data management platform and product

Abstract

The invention discloses a multi-center clinical research data management platform and product, relates to the field of data management, and is used for improving the sharing convenience of multi-center clinical research data, reducing the project recruitment threshold and improving the shared data quality. The platform comprises a registration module used for creating a user account; the reference library module is used for constructing and managing a reference library; the clinical research module is used for initiating and managing clinical research items; and the integral module is used for constructing an integral rule and managing the integral of each user in the clinical research project according to the integral rule. According to the invention, a multi-center clinical research data sharing platform for online communication and cooperation is provided for numerous researchers, the convenience of project recruitment is improved, the resource utilization rate of branch centers is improved, the patient data entry efficiency and objective accuracy are improved, and the multi-center cooperation enthusiasm is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data management, and in particular to a multi-center clinical research data management platform and product. BACKGROUND

[0002] Most existing data management platforms only manage data for a single center / agency and do not have functions such as data statistics, analysis, and sharing, which is not conducive to the development of multi-center clinical research in the big data era.

[0003] In addition, since many existing data management platforms do not have the function of allocating points according to the quality and quantity of enrolled patients after data sharing, the value brought by data sharing cannot be reflected, and researchers are unwilling to share their data, resulting in the development of multi-center clinical research having to be carried out after private negotiation and signing of contracts, which is complex, time-consuming, and labor-intensive, and has high time and material costs for project recruitment. In addition, existing data management platforms do not have a follow-up reminder function, which leads to data follow-up often being overdue or even lost, affecting the quality of the data. SUMMARY

[0004] The present application aims to: in view of all or part of the above-mentioned problems, provide a multi-center clinical research data management platform and product to improve the convenience of multi-center clinical research data sharing, reduce the project recruitment threshold, and improve the quality of shared data.

[0005] The technical solutions adopted by the present application are as follows: A multi-center clinical research data management platform, comprising: A registration module configured to receive a user registration request and create a user account.

[0006] A benchmark library module configured to build and manage a benchmark library; wherein the work of building the benchmark library includes building at least one CRF form under each benchmark library; the work of managing the benchmark library includes selecting multiple basic components for combination for each CRF form in a pre-built basic component set, and entering patient data; A clinical research module configured to initiate and manage a clinical research project; wherein initiating the clinical research project includes newly creating a clinical research project for a main center and adding a sub-center to the newly created clinical research project; managing the clinical research project includes receiving newly added patient data from the main center and the sub-center, and analyzing the patient data of the clinical research project; A point module configured to build a point rule and manage the points of each user in the clinical research project according to the point rule.

[0007] A computer program product comprises a computer program, which, when executed by a processor, executes the multi-center clinical research data management platform described above.

[0008] Therefore, by adopting the technical scheme described above, the present application has the beneficial effects that: The multi-center clinical research data management platform provided by the present application has multi-center clinical research project initiation and participation functions, and provides an online communication and cooperation multi-center clinical research data sharing platform for a plurality of researchers. Different researchers (users) can initiate or find clinical research projects of their own interest (breast disease) online, which can effectively avoid repeated research, and gather multi-center resources to complete projects that cannot be undertaken by a single center, improve the utilization rate of sub-center resources, and reduce the time and material costs spent on the recruitment process. The automatic follow-up and data supervision functions of the platform also ensure the quality of the clinical research and the reliability of the data, saving the process supervision cost. The platform also designs data statistics and analysis functions, and users can view the research results of the project at any time, conveniently and quickly understand the project status, and quickly form written results. The credit module of the platform can accurately assign the contribution degree of each center to the project as a value basis for the output of research results, which can not only improve the enthusiasm of researchers participating in research, but also serve as a retrospective function of the historical research participation degree of researchers, to objectively reflect the contribution of different researchers to scientific research results. BRIEF DESCRIPTION OF DRAWINGS

[0009] The present application will be described by way of example and with reference to the accompanying drawings, in which: Figure 1 is a structural diagram of the multi-center clinical research data management platform.

[0010] Figure 2 is a registration module display interface.

[0011] Figure 3 is a reference library module structural diagram.

[0012] Figure 4 is a constructed reference library schematic diagram.

[0013] Figure 5 is a basic component set schematic diagram.

[0014] Figure 6 is a basic library patient data entry schematic diagram.

[0015] Figure 7 is a chart search result schematic diagram.

[0016] Figure 8 is a pre-entry method flowchart.

[0017] Figure 9is a clinical research project initiation flowchart.

[0018] Figure 10 is a recruitment interface diagram.

[0019] Figure 11 is an enrollment application interface diagram.

[0020] Figure 12 is an enrollment application flowchart. DETAILED DESCRIPTION

[0021] All features disclosed in this specification, or all steps of any methods or processes disclosed, can be combined in any combination, except combinations where at least some of such features and / or steps are mutually exclusive.

[0022] Any feature disclosed in this specification, unless stated otherwise, can be replaced by any equivalent or similar feature, or by a feature that serves the same, similar or equivalent purpose. That is, each feature disclosed in this specification is one example only of a range of equivalent or similar features.

[0023] In view of the fact that existing single-center data management platforms are only responsible for managing clinical research data of a single center and cannot share and cooperate online, and the fact that complex clinical research projects can only be recruited offline, which is time-consuming and laborious, embodiments of the present application provide a multi-center clinical research data management platform and product, which aims to improve the convenience of sharing multi-center clinical research data, reduce the difficulty of project recruitment, improve the quality of shared data, and improve the reliability of project research results.

[0024] Embodiments of the present application provide a multi-center clinical research data management platform, as shown in Figure 1 , which comprises: The registration module is configured to receive a user registration request and create a user account.

[0025] As shown in Figure 2As shown, the user registration request contains the basic information filled in by the user when registering a user account on the platform, such as name, contact information, ID number, title / identity (such as medical assistant, physician, attending physician, deputy director, director), department, hospital, affiliated medical group (main diagnosis group), whether it is a medical group leader, password, etc. Among them, "whether it is a medical group leader" determines whether the user account has the right to initiate a clinical research project, and determines the rights of other members participating in the clinical research project to operate the clinical research data. In the created user account, the user is divided into two types: the main center as the medical group leader and the branch center as the non-medical group leader. (The created account is the smallest unit of the medical group. There may be multiple medical groups in the same center (hospital), that is, each center can register multiple different accounts. In addition to the medical group leader, there are other members in a medical group, only the permission allocation is different. The multi-center research project is who initiates the clinical research project, and the medical group is the main center of the clinical research project. The medical group participating in the clinical research project is the branch center.) The participants of each clinical research project are a medical group, and the medical group is the smallest unit of the project. Usually, in addition to the medical group leader, the other members in a medical group only have the right to enter clinical research data (patient data).

[0026] The user registration request is audited by the platform administrator, and after the audit is passed, the user account is successfully created.

[0027] The benchmark library module is configured to build a benchmark library. The work of building a benchmark library includes building at least one CRF form under each benchmark library.

[0028] For example, as shown in Figure 3 The built benchmark library includes "benchmark library", "prospective benchmark library", "retrospective benchmark library", "other benchmark library", etc. Taking "benchmark library" as an example, "benchmark library" studies breast diseases, as shown in Figure 4 The CRF form built under it includes "preoperative basic information", "preoperative breast shape characteristics", "preoperative breast tumor characteristics", "intraoperative data", "postoperative follow-up surgical complications", etc.

[0029] The benchmark library module is further configured to manage the benchmark library. The work of managing the benchmark library includes selecting multiple basic components for each CRF form for combination in the pre-built basic component (field) set.

[0030] For example, as shown in Figure 5As shown, the pre-constructed basic component set includes "single-line text", "integer input box", "date and time", "drop-down single selection", "file upload", etc. By configuring the properties of each CRF form (configured through the "CRF property setting" menu), multiple basic components are selected for each CRF form, and the properties of each selected basic component are configured separately (configured through the "control property setting" menu). For example: basic component name (such as name, gender, age, etc.), item code (i.e. basic component identification code), placeholder content, whether to hide, hidden trigger condition, whether to fill in, etc. This function can greatly shorten the tedious configuration time of the CRF form, and provide protection for the user's data import. In addition, users can set their own image sets (classified by time points such as preoperative, one week after operation, one month after operation, etc.) in the CRF form configuration interface, which can be named by the user and can be exported and downloaded.

[0031] For each reference library, according to the configured patient basic information control, patient data can be added to each reference library by adding or importing patient basic information. For example, Figure 6 As shown, the patient data is, for example, "patient number", "patient name", "study type", etc. Each patient data is automatically configured with all the CRF forms of the reference library to which it belongs, so as to facilitate the perfection of clinical research data. In addition, each CRF form is configured with a state indication mark to indicate the submission state of the corresponding CRF form, such as "form submitted", "form not completed", "form not started", etc.

[0032] Through the above-mentioned reference library module, the reference library can be created conveniently and quickly. Users can create the required reference library efficiently and flexibly according to the platform guide, and users can achieve zero threshold library building according to the platform guide, and can obtain good user experience.

[0033] In addition, users can configure the follow-up plan (such as follow-up trigger condition, follow-up time interval, follow-up cutoff condition, etc.) for each patient data in each reference library, which facilitates users to regularly follow up their patients and realizes automatic reminders.

[0034] For patient data (including CRF forms) under each reference library, relevant patient data can be searched through key words, patient-affiliated institutions, creation time, etc. The searched patient data can be exported in single or batch. In addition, the analysis results of the searched patient data can also be exported. For example, Figure 7 As shown.

[0035] For retrospective study data and past patient data (usually recorded in Excel format), the user can import the patient data in one key after keeping the relevant table header content consistent with the benchmark library table header content, greatly improving the efficiency of creating and perfecting the basic library, and also avoiding the operation errors that may be introduced by manual entry. The data that cannot be recognized during import will be specially marked for easy modification or manual input.

[0036] The patient information entered into the benchmark library can be matched by the pre-configured pre-entry group method to determine whether the patient data meets the current clinical research project that the user is participating in, and a pre-entry group clinical research list can be generated. When the patient data meets the requirements of any clinical research project that the user is participating in, the pre-entry group clinical research list is automatically popped up for the clinical research project. A corresponding pre-entry group clinical research list is generated for each clinical research project that the user joins. The platform receives the user's confirmation information for entering the pre-entry group clinical research list into the clinical research project, and automatically synchronizes the patient data to the corresponding clinical research project. In addition, if there is a conflict between the clinical research projects that the patient data can enter, the patient information is synchronized to the clinical research project that the user finally determines according to the user's selection. In addition, after the patient data in the benchmark library is updated, the corresponding patient data in each clinical research project is automatically updated.

[0037] Through the innovative pre-entry group method designed by the platform, intelligent recognition and rapid entry of patient data from the benchmark library to the clinical research project are realized, which greatly improves the efficiency of patient data entry and the value of patient data, and also eliminates the subjective errors that may be introduced by the user's subjective selection of patient data (such as entering incorrect patient data or missing compliant patient data), thereby improving the reliability of the research results of the clinical research project.

[0038] As an optional implementation, the above-mentioned pre-entry group method is based on the concept of medical ontology driving and multi-modal feature fusion, which realizes precise and efficient patient matching between patient data and the entry criteria set for the clinical research project (such as age, gender, feature value range, etc., which are set when the user creates the clinical research project) by combining clinical rules with machine learning technology and adopting a four-layer intelligent matching architecture. When the entry criteria are set in natural language, they are first parsed into machine-understandable logical expressions, for example, time-related conditions are parsed into computable time constraint networks.

[0039] The four-layer intelligent matching architecture includes a data preprocessing layer, a feature calculation layer, an intelligent matching layer, and a result optimization layer. This architecture supports automatic screening of target patient groups that meet specific clinical research projects from a multi-center breast disease patient benchmark library.

[0040] Specifically, as shown in Figure 8 , it includes: I. Data preprocessing layer.

[0041] The data preprocessing layer is responsible for the unified standardization of patient data from different centers (clinical research project participants). The data preprocessing layer performs the following operations: Establish a breast disease-specific medical terminology ontology library to achieve unified mapping of multiple standard diagnostic codes such as ICD-10 and SNOMED CT; Design an automatic conversion mechanism for index units to eliminate differences in detection standards among centers; Use time series data alignment algorithms to unify data records at different follow-up time points.

[0042] II. Feature calculation layer.

[0043] Patient data features are divided into five core dimensions, each containing multiple sub-features: (1) Demographic dimension: basic information such as age, BMI, menstrual status, and reproductive history; (2) Clinical pathology dimension: tumor stage, size, lymph node status, histological grade, etc.; (3) Molecular typing dimension: ER / PR / HER2 status, Ki-67 index, gene expression profile, etc.; (4) Treatment response dimension: neoadjuvant therapy effect, disease-free survival, adverse reactions, etc.; (5) Time series dynamic dimension: disease progression pattern, treatment sequence, biomarker trajectory, etc.

[0044] According to the hard conditions set for different features in the inclusion criteria, exclude patient data that does not meet the hard conditions. For different types of clinical research projects, the hard conditions set for different features differ in threshold values or priority levels. Therefore, first match the corresponding hard conditions according to the type of clinical research project, and then filter / select patient data according to the matched hard conditions.

[0045] In addition, in the feature calculation layer, a feature credibility scoring mechanism is established to assign quality weights to each dimension feature based on patient data integrity, data standardization degree, and time consistency, to weight the matching degree of each dimension feature, and ensure the reliability of the pre-inclusion method.

[0046] III. Intelligent matching layer.

[0047] The intelligent matching layer uses a multi-granularity attention matching mechanism, a medical ontology enhanced semantic matching mechanism, and a time series dynamic feature matching mechanism to match patient data and inclusion criteria.

[0048] Smart matching layer is to match patient data (including disease diagnosis, genetic characteristics, medical history, etc.) with the inclusion criteria of clinical research projects to identify patients who meet the conditions of specific clinical trials. The matching object is: patients and the inclusion criteria of clinical research projects.

[0049] (1) Multi-granularity attention matching mechanism.

[0050] The multi-granularity attention matching mechanism adopts a three-level matching architecture: 1. Feature-level matching (fine-grained): precise matching for each sub-feature.

[0051] According to the type of each sub-feature, different methods are used for matching: For numerical features, an adaptive matching algorithm is used for matching; For categorical features, semantic similarity is calculated based on medical ontology for matching; For time series features, dynamic time warping algorithm is used for matching.

[0052] As an optional implementation, feature weight templates for different types of clinical research projects are pre-constructed. For example, clinical research projects can be divided into "efficacy evaluation", "prognosis research", "biomarker exploration", etc. Each type has a corresponding feature weight template to indicate the weight of each sub-feature, representing the importance of the sub-feature in the current clinical research project. The matching result (matching degree) of each feature is weighted using the weight indicated by the feature weight template corresponding to its type. The pre-constructed feature weight template can be manually adjusted by the user.

[0053] 2. Dimension-level matching (medium-grained): calculate the importance weight of each dimension feature at the feature set level.

[0054] Considering the clinical relevance constraints between dimension features (i.e. only considering patient data that meets the constraints), an attention mechanism is applied to automatically learn the importance weight of each dimension feature. Among them, the attention network is trained through historical matching data, so that the algorithm can automatically identify the most important feature combination for a specific clinical research project, and realize adaptive optimization of the matching strategy.

[0055] 3. Patient-level matching (coarse-grained): comprehensive evaluation of the overall compliance of patient data.

[0056] The weight distribution is adjusted by introducing clinical expert prior knowledge. For example, according to the weight coefficient of the clinical expert prior knowledge and the weight coefficient of the feature set, the weight of each dimension feature calculated based on the clinical expert prior knowledge is weighted and summed with the weight of each dimension feature obtained by the dimension-level matching to obtain the final weight of each dimension feature. According to the final weight of each dimension feature, the matching degrees of each dimension feature are weighted and fused to obtain the matching degree under the multi-granularity attention matching mechanism.

[0057] According to the multi-granularity attention matching mechanism, the patient data is sorted in descending order of matching degree, and the patient data whose matching degree does not reach the matching degree threshold is generated into a candidate patient list.

[0058] As an optional implementation, the medical ontology enhanced semantic matching mechanism is used to verify and optimize the candidate patient list.

[0059] (2) Medical ontology enhanced semantic matching mechanism.

[0060] The medical ontology enhanced semantic matching mechanism also adopts a three-level matching architecture: 1. Breast disease ontology construction.

[0061] The breast disease ontology construction includes the following operations: Establish a hierarchical ontology structure containing concepts such as diagnosis, treatment, pathology, and genes; Define semantic relationships between concepts (such as "is_a", "part_of", "treated_by", etc.).

[0062] 2. Semantic similarity calculation.

[0063] The semantic similarity calculation includes the following operations: Calculate the similarity of diagnostic terms based on ontology path distance, wherein the ontology path distance is calculated based on the concept information content and clinical relevance in the diagnostic term to obtain a weighted result.

[0064] Exclude patient data whose similarity does not reach the set threshold.

[0065] In addition, the calculation of semantic similarity also supports fuzzy matching and concept expansion to improve recall rate.

[0066] 3. Treatment path matching.

[0067] The treatment path matching includes the following operations: The patient data is analyzed for logical consistency with the clinical research project. In performing the logical consistency analysis, the degree of influence of dynamic factors such as treatment timing and dose adjustment is considered (conversion of factor consistency according to the preset influence of the factor). In addition, the value of alternative treatment options in the clinical research project is also considered, and the logical consistency of the alternative treatment options with the clinical research project is converted according to the conversion relationship between the value and the logical consistency. Patient data that does not meet the set threshold for logical consistency is excluded.

[0068] As an optional implementation, for the screened candidate patient list, a time sequence dynamic feature matching mechanism is used for verification and optimization.

[0069] (3) Time sequence dynamic feature matching mechanism.

[0070] The time sequence dynamic feature matching mechanism uses a two-level matching architecture: 1. Time series analysis.

[0071] Time series analysis includes the following operations: Extract key time point features of disease progression; analyze the consistency of key time point biomarker change trends with the clinical research project. The patient data for time series analysis must meet the time constraints such as treatment interval and follow-up period. Patient data that does not meet the set threshold for consistency is excluded.

[0072] 2. Dynamic inclusion and exclusion criteria support analysis.

[0073] The calculation of dynamic inclusion and exclusion criteria support includes the following operations: Processing time-related conditions such as "within 6 months of diagnosis" and "3 months after treatment ends"; Supporting patient stratification matching based on disease progression speed; Adapting to the timing logic requirements in the clinical research project.

[0074] Patient data that does not meet the set threshold for support is excluded.

[0075] After one or more of the above matching mechanisms, the pre-entry clinical research list is obtained.

[0076] The dynamic feature matching process includes: (1) First step: Time-related condition analysis and standardization.

[0077] Extract time expressions: Extract all time-related conditions from the clinical trial entry criteria, such as "within 6 months of diagnosis" and "3 months after treatment ends".

[0078] Semantic analysis: Use natural language processing technology (NLP) to analyze time expressions, identify time reference points (diagnosis, treatment end, etc.), and time thresholds (6 months, 3 months, etc.).

[0079] Standardized conversion: Parse the time expression into a rule form combined with the corresponding time threshold. For example: Convert "within 6 months of diagnosis" to: diagnosis date + 6 months > current date; Convert "3 months after treatment ends" to: treatment end date + 3 months < current date; Convert "follow up once every 3 months" to: follow-up interval = 3 months.

[0080] Build a time rule library: store the parsed rules in the medical knowledge graph for subsequent matching.

[0081] (2) Second step: patient time feature extraction and quantification.

[0082] Time data extraction: extract the patient's key time points from the electronic medical record system: diagnosis date, treatment start date, treatment end date, last follow-up date, etc.; extract time-related descriptions from medical records, such as "3 months after diagnosis", "2 months after treatment ends", etc.

[0083] Time feature calculation: Calculate the time difference from diagnosis to date: diagnosis to date = current date - diagnosis date; Calculate the time difference from treatment end to date: treatment end to date = current date - treatment end date; Calculate the time difference from last follow-up to date: last follow-up to date = current date - last follow-up date.

[0084] Time feature standardization: Convert all time differences to days (e.g. 6 months = 180 days); Add time unit (e.g. days, months, years) identifiers to each time feature.

[0085] (3) Third step: disease progression speed stratification matching.

[0086] Disease progression indicator extraction: Extract multiple time points (e.g. each diagnosis or follow-up) of disease indicators (e.g. tumor size, biomarker level, symptom score, etc.) from medical records; identify key time points of indicator changes.

[0087] Progression speed calculation: Calculate the change rate of indicators at different time points: change rate = (indicator value 2 - indicator value 1) / (time difference); indicator value 2 is the disease indicator at the later time point, and indicator value 1 is the disease indicator at the earlier time point.

[0088] Stratification standard establishment: Based on the disease progression criteria in the medical knowledge graph, patients are divided into different levels: Fast progression level: rate of change > 0.5% per day; Medium progression level: 0.1% ≤ rate of change ≤ 0.5% per day; Slow progression level: rate of change < 0.1% per day; Set the corresponding disease progression speed threshold for each level.

[0089] (4) Fourth step: Time sequence logic matching and weight calculation.

[0090] Time sequence rule matching: Match the patient's time characteristics with the time sequence requirements in the clinical trial enrollment criteria.

[0091] Matching degree calculation:Calculate the matching degree for each time condition: Matching degree = 1 - | Actual time - Threshold | / Threshold. Time sequence logic weight assignment:Assign weights to different time conditions according to the importance of clinical trial design:

[0092] Enrollment time requirement (within 6 months after diagnosis): weight 0.4; Treatment end waiting period (3 months after treatment end): weight 0.3; Follow-up interval requirement (follow-up every 3 months): weight 0.2; Other time-related conditions: weight 0.1; Calculate the comprehensive time sequence matching degree: Comprehensive matching degree = Σ (Matching degree_i × Weight_i). (5) Fifth step: Multi-granularity attention matching and result optimization.

[0093] Multi-granularity attention mechanism: Apply attention mechanism on time characteristics to automatically learn the importance of different time scales. Here, "multi-granularity" represents different levels of time scale features, such as hour-level, day-level, month-level time window features, or other time sequence patterns under different time units. Through attention mechanism, different scale time features are assigned weights to focus on multi-scale time sequence information.

[0094] Medical ontology enhanced matching:Use the disease progression relationship in the medical knowledge graph to enhance the semantic accuracy of matching.

[0095]

[0096] The disease progression relationship in the medical knowledge graph (such as disease A often occurs before disease B, disease C is associated with symptom D, etc.) is used to integrate medical semantic logic in the matching process, thereby enhancing the semantic accuracy of matching. For example, when the patient's medical history contains disease A, the subsequent progression path of disease A in the medical indication graph is combined to enhance the semantic accuracy of matching the disease development timing, associated symptoms, etc., avoiding matching deviations that are out of sequence characteristics and rely only on surface word features.

[0097] Timing dynamic feature fusion: The time feature and other clinical features of the patient (such as disease type, age, gender, etc.) are fused through a feature interaction layer, which associates the time feature and the clinical feature. The fused feature contains both the time dimension information and the clinical attribute logic. Through the "dynamic feature matching mechanism", the matching result is ensured to comply with the timing logic. This mechanism dynamically adjusts the matching logic according to the changes in the patient's timing characteristics, i.e. the matching rules for patient data and enrollment rules are different at different time steps, allowing the model to learn the weight of different time step features and achieve adaptive matching of timing rules.

[0098] Through the dynamic feature matching mechanism, the matching result is ensured to comply with the timing logic.

[0099] Final matching sorting: Sort patients according to the comprehensive matching degree; Prioritize patients who meet the matching degree threshold; Output the matching result, including the matching degree score and specific matching basis.

[0100] Four, result optimization layer.

[0101] The result optimization layer is used to refer to the matched pre-enrollment clinical research list to feedback the feature weight of the feature extraction stage and the threshold in the feature matching process. For example, optimize the feature weight template and set the threshold.

[0102] Through the above pre-enrollment method, the intelligence and efficiency of patient data entry into clinical research projects can be greatly improved, and the workload and subjective errors of manual entry can be reduced.

[0103] The clinical research module is configured to initiate clinical research projects. Initiating a clinical research project includes creating a new clinical research project for the primary center and adding sub-centers to a newly created clinical research project. Only user accounts with the role of Medical Team Leader (i.e., the medical team acting as the leader) can initiate research requests and create new clinical research projects. User accounts without the role of Medical Team Leader (i.e., sub-centers) can choose whether to join a newly created clinical research project during the initiation phase. Sub-centers refer to other medical team accounts, besides the primary center, that participate in the clinical research project and are led by the Medical Team Leader. The terms "primary center" and "sub-center" are relative; a user may act as the primary center in one clinical research project but simultaneously as a sub-center user in other clinical research projects.

[0104] like Figure 9 As shown, when the main center initiates a research request, the platform requires the medical team leader to upload a research summary of the clinical research project (including research name, center type: multi-center or single-center study, research type: which type of clinical research it belongs to, research objective, brief description of the research protocol, inclusion criteria, exclusion criteria, requirements for sub-centers, number of sub-centers to be recruited, research recruitment start and end dates, research start and end dates, estimated publication level, estimated number of publications, project leader's professional resume, project leader's contact information, project leader's mailing address, funding or material source, etc.). After the research summary passes the initial review, sub-center recruitment begins. After sub-center recruitment is completed, the medical team leader initiating the clinical research project needs to upload detailed information about the research again for a second review. Due to compliance requirements related to the research, no operations can be performed on the clinical research project before the second review is passed, such as patient enrollment or changes to the research's CRF form. Single-center studies (i.e., 0 sub-centers recruited) do not have a sub-center recruitment process.

[0105] For multi-center studies, the medical team leader can decide the number of sub-centers to recruit when uploading the study summary. After the medical team leader completes the initial review of the study, the relevant study will be displayed on the homepage or the "Studies I've Participated In" page. Medical teams at sub-centers interested in the clinical research project can apply to join. Once a sub-center is fully recruited, other sub-centers cannot join, and recruited sub-centers cannot withdraw from a clinical research project they have already joined. If other sub-centers are allowed to join a relevant clinical research project after it has entered the research process, the main center can also initiate a second recruitment. Clinical research projects that conduct a second recruitment must re-upload relevant documents such as ethics approval documents for approval. Sub-centers only have the authority to enter patient data in relevant clinical research projects and cannot make any changes to the configuration of the clinical research project.

[0106] As an alternative implementation method, the medical team leader can proactively invite sub-centers to join the clinical research project.

[0107] The initiation process of the above clinical research project can greatly save the time and cost of building and recruiting the clinical research project.

[0108] The medical group leader modifies the research status of the clinical research project to "in progress" through the clinical research module after the establishment of the clinical research project (i.e., the review is passed). Therefore, the clinical research module is also configured to manage the clinical research project. The work of managing the clinical research project includes receiving newly added patient data of the user and analyzing the patient data of the clinical research project.

[0109] The main center has the highest authority for this clinical research project and can configure the project information as a whole: project configuration, CRF form management, and research status change: changing the research status to termination, completion, and deleting the project (after being reviewed by the background management personnel, the clinical research project is deleted). The following focuses on the project configuration and CRF form management functions.

[0110] (1) Project configuration.

[0111] In the main interface of initiating the research, the research initiator can perform overall project configuration on the research, including research information, center and personnel configuration, demographic information, automatic coding, automatic follow-up, randomization, etc.

[0112] Research information: relevant information filled in when initiating the research request, including the research profile and uploaded files. If changes are to be made to this item, it needs to be submitted to the background management personnel for review.

[0113] Center and personnel configuration: permission management for sub-centers participating in the clinical research project and user accounts of the main center except the medical group leader. Any sub-center participating in the clinical research project only has the permission to enter patient data. The medical group leader can assign roles and configure permissions for other personnel in his center (i.e., the main center). The configuration granularity is accurate to each CRF form of the research, and there are four types of permissions for configuration: adding, saving, submitting, and deleting. For example, in a clinical research project named benign mass, there are four CRF forms. If the role of Zhang San in the main center is assigned as the data entry person for the first CRF form, then Zhang San only has the permission to add and save the first CRF form in this clinical research project, and Zhang San cannot perform any operation on the remaining three CRF forms.

[0114] Demographic information: basic information that patients enrolled in the study need to fill in, including patient number, name, gender, ID number, date of birth, age, phone number. The medical group leader can also change the information to be filled in according to the needs and can also choose whether to enable the relevant information and whether it is mandatory.

[0115] Automatic coding: coding for enrolled patients. The medical group leader can choose the configuration type of coding according to the needs.

[0116] Auto Follow-up: The medical group leader can open the auto follow-up function of the clinical research project. After opening the auto follow-up function, the prepared follow-up plan is assigned to each patient in the clinical research project through the key field triggered by the follow-up. The medical group leader and the sub-center can configure multiple follow-up plans. The follow-up plan configured by the medical group leader is applicable to the entire clinical research project, and the follow-up plan configured by the sub-center is only applicable to the sub-center.

[0117] Randomization: It refers to a method of completely determining patient data by chance (like drawing lots) and randomly assigning them to different research groups in a clinical research project. Randomization is configured according to the needs of different clinical research projects, including: 1) Group configuration: Blind selection, including "non-blind", "single-blind", "double-blind". The sample size of the clinical research project needs to be filled in, and the group information needs to be set.

[0118] 2) Random plan configuration: Random methods can choose "simple random", "block random", "stratified block random". If "block random" is selected, "block length" needs to be filled in; if "stratified block random" is selected, "block length", "stratified institution", and "stratified factor" need to be filled in.

[0119] 3) Random number configuration: Fill in "prefix", "project serial number", and "suffix". Once the random plan is generated and a patient is enrolled in the clinical research project, the configuration cannot be changed. In a double-blind trial, the "unblinding" can also be performed at a specific time of the clinical research. "Randomization" strictly follows the plan prepared in advance in the clinical research project.

[0120] (2) CRF form management.

[0121] The medical group leader configures the CRF form required by the research institute in the same way as adding a CRF form in the benchmark library, including adding, importing, exporting, and deleting the CRF form required by the research institute. At the same time, the newly added CRF form is also composed of multiple basic components added from the pre-built basic component set. The CRF form in the project can be configured for attribute control, component naming, and trigger condition configuration. The CRF form in the project and the CRF form in the benchmark library do not interfere with each other and are independent of each other. They can also import or export patient data.

[0122] The CRF form in the benchmark library or the clinical research project is standardized through a standard and normative basic component set, ensuring the convenience of creating CRF forms while forming a unified and scientific standard, ensuring the portability and scientificity of patient data.

[0123] After the research status of the clinical research project is changed to "research in progress", the clinical research module displays the overall research situation on the main interface of the clinical research project, including the number of patients enrolled in the project (multiple instances of not starting data entry, multiple instances of data entry in progress, and multiple instances of completed data entry), the number of patients enrolled in the field (age statistics, gender statistics, and operation statistics, etc. according to the statistical fields set by the medical group leader in this project), the total number of patients enrolled, the number of patients enrolled per month, the proportion of institutions enrolled, the operation record, etc. Through the data supervision function of the platform, the quality of the clinical research and the reliability of the data are ensured.

[0124] The main interface of the project is a specific embodiment of research data entry. New patients in the project can be added through two ways: adding new patients and importing data. The demographic information of the new patient needs to be filled in first, and after filling in, the patient's CRF form filling situation can be viewed on the main interface. The medical group leader (or project participant) needs to fill in the patient data in the Excel table, align the Excel table header information with the CRF form header content set by the clinical research project, and realize one-key import to achieve basic data import.

[0125] In addition, the medical group leader or project participant can migrate patient data from their respective reference library to the clinical research project they participate in with one key, or vice versa, migrate patient data already entered into the clinical research project they participate in to their respective reference library, avoiding repeated patient data entry, greatly saving the time of patient data entry.

[0126] The medical group leader can accurately locate the patient through key fields, affiliated institutions, and creation dates. The patient's basic information, CRF form submission situation, and reference library are the same, and are also specifically embodied on the main interface.

[0127] After clicking the follow-up plan on the main interface of the project, the medical group leader or sub-center can add a follow-up plan for the clinical research project by filling in the basic information of the follow-up plan.

[0128] The follow-up configuration has two options: fixed interval and custom. First, complete the trigger condition configuration of the follow-up plan (for example: complete the operation date and time filling in a certain CRF form), the medical group leader / sub-center can set the follow-up time, reminder time, and push time after the follow-up is triggered, that is, how many days after the trigger to follow up, how many days in advance to remind the patient, and complete the follow-up push. The types of follow-up push include CRF form push and text push, and the receiving personnel (receiving type) include patients, doctors, doctors and patients, and multiple options.

[0129] In addition to adding follow-up plans in the patient list, the medical group leader can also set automatic follow-up in the project configuration. All patients participating in the study must be followed up. In the launched study, once the follow-up plan set by the study initiator of the main center (i.e. the center where the medical group leader is located) is configured, it must be completed by both the main center and the branch center. The follow-up plan set by the branch center is only used in that center.

[0130] Patient data in the clinical research center can also be retrieved and exported, in the same way as patient data retrieval and export in the reference library. There are two export modes for patient data, one is Excel table export (single patient data or batch export), and the other is statistical chart export.

[0131] Excel table export: Users can select the Excel table export range (reference library-CRF form-table content), sheet type (multiple sheets: all export content in one large table, single sheet: export content according to CRF form), and data format (vertical, horizontal) by themselves. You can also limit the data content to be exported by key fields to achieve precise positioning.

[0132] Statistical chart export: Similar to Excel table export function, users can select the export range by themselves, search by key fields, and limit the export content of each chart, such as the male-female ratio in open breast surgery. The export format has a line chart, with a line chart, pie chart, column chart, ring chart, box chart, column line Figure Six This function can facilitate users to perform research data statistics, work report, and research explanation, etc.

[0133] In the calendar interface of the clinical research project, the medical group leader can set specific events on specific dates. There are four types of events to choose from: personal items, meetings, team items, and follow-ups (for example: set a team event on a certain day in a certain month of a certain year, and hold a seminar for the whole team). Among them, in addition to the follow-up events added by the user, after setting the follow-up plan for the patient in the patient list, a specific event will also be generated on the specific date in the calendar interface.

[0134] In addition, the platform can also be associated with WeChat mini programs to achieve related information push, such as follow-up content or edited text messages that need to be filled in by patients. At the same time, the platform generates a patient-specific two-dimensional code, which can be scanned by patients during re-examination to view the patient's relevant information filling situation.

[0135] Currently in the medical field, patient follow-up is mostly by the physician to schedule the patient for a return visit at an agreed time, and the follow-up information is obtained by filling out a paper CRF form. As a major improvement of the present application platform, in the embodiments of the present application, the platform generates a special two-dimensional code for each patient and associates it with a mini-program / APP, and according to the determined follow-up time, the platform pushes a follow-up reminder to each patient on the follow-up date through the mini-program / APP, and after obtaining confirmation, the platform automatically generates a follow-up list for the follow-up date. When the patient is followed up, the patient scans the special two-dimensional code and fills in the follow-up information (such as filling in the relevant examination indicators when returning for a visit) in the CRF form. The doctor (i.e. the user of the platform) scans the special two-dimensional code of the patient (generated by the platform on the patient data page) to view the missing follow-up information of the patient, and by clicking on the perfect data, the doctor automatically jumps to the perfect interface of the missing follow-up information, and after perfecting, the doctor automatically jumps to the next missing follow-up information, and so on, until all the missing follow-up information is displayed. In the perfect interface of the follow-up information, the reason for not collecting the condition can be obtained. The platform automatically calculates the relevant physiological indicators according to the CRF form filled out by the patient, and warns the physiological indicators that do not meet the standard. For example, the doctor sets a warning threshold for the physiological indicators of the clinical research project, and the platform filters out the patient data whose physiological indicators do not meet the warning threshold according to the warning threshold, and highlights the items that do not meet the standard (such as red).

[0136] Through the above follow-up scheme design, the platform can remind the patient in time, and automatically plan the follow-up period (i.e. sequencing) for the patient, which can avoid missing reminders and facilitate the patient's time arrangement. In addition, the patient can independently enter the follow-up information through the special two-dimensional code, which eliminates the restrictions of fixed filling location, fixed filling method, and the need to scan and enter the system for index evaluation of the paper CRF form, and can improve the patient's follow-up experience; the doctor can easily check whether the follow-up information of each patient is missing, so as to timely check and make up for the missing information, avoid the problem of data missing that cannot be made up due to lag in collecting paper CRF forms, and reduce the work burden of the patient and the doctor. The platform can automatically calculate the corresponding physiological indicators after the follow-up information is entered, and give a warning, which can timely discover the risk factors and protect the safety of the patient. Through the association of the WeChat / APP, the platform can also push relevant medical popular science, health management and other information to the patient, and improve the patient's experience and compliance.

[0137] As shown in Figure 10 , the sub-center (i.e. the user who does not initiate a research request) browses the clinical research projects initiated by others on the recruitment interface of the platform, or searches for the clinical research projects of interest on the recruitment interface through project keywords (such as project name, research type, initiating agency, etc.). As shown in Figure 11As shown, after selecting a clinical research project of interest, detailed information of the selected clinical research project can be viewed on an introduction interface of the clinical research project. In addition, an enrollment application interface is provided on the introduction interface of the clinical research project. After clicking the enrollment application interface and submitting necessary application materials, an enrollment application can be initiated. The enrollment application is audited by the main center of the applied clinical research project. Only after the main center passes the audit, the sub-center can join the clinical research project.

[0138] Specifically, as shown, Figure 12 the audit of the main center on the enrollment application includes a preliminary review and a review, wherein the preliminary review includes the proof materials uploaded by the sub-center to prove the compliance with the research requirements, such as historical research direction, education, qualification, historical project, historical research achievement, etc. After the main center passes the audit on the proof materials and the ethics review, the sub-center can download the materials such as the ethics approval and the research plan of the clinical research project. Thereafter, the sub-center uploads the materials such as the ethics approval, which are reviewed by the main center. After the main center passes the review, the platform background management personnel performs an audit, and the audit content includes, for example, the completeness of the project materials and the qualification of the project participants. After the management personnel passes the audit, the sub-center successfully joins the applied clinical research project. During the research, the sub-center is responsible for adding patient data. Once the sub-center adds patient data, the sub-center cannot modify or delete the data. If modification or deletion is needed, the sub-center needs to apply to the main center or the management personnel separately.

[0139] The credit module is configured to build a credit rule and manage the credit of each user account in the clinical research project according to the credit rule.

[0140] The credit rule is a credit increase and decrease rule configured according to the type of the joined clinical research project, the work completed in the clinical research project, and the credit consumption of the user, which can be flexibly designed according to specific application scenarios and specific research work.

[0141] For example, as an optional implementation, the credit rule is configured as: (1) Credit acquisition rule.

[0142] 1 case of enrolled patient credit = journal impact factor x journal partition article score / sample size x 100 (rounded to the nearest whole number). According to the estimated 1 case of enrolled patient credit and the number of effective patient data of each center, the credit is distributed to each center.

[0143] Article partition: 24 article points for Zone 1, 12 article points for Zone 2, 8 article points for Zone 3 (baseline library prospective), 6 article points for Zone 4, 4 article points for Zone 5, and 2 article points for Zone 6 (baseline library retrospective).

[0144] (2) Credit consumption rule.

[0145] The total score of an article is defined as 10 points: Zone 1 / Zone 2: The first author and the first corresponding author each account for 1 point, the second co-first author and co-corresponding author each account for 3 points, and the third co-first author accounts for 2 points; Zone 3 / Zone 4: The first author and the corresponding author each account for 2.5 points, and the second co-first author accounts for 5 points; Zone 5 / Zone 6: The first author and the corresponding author each account for 5 points.

[0146] Author Quantity Limitations: Zone 1 / Zone 2: First author and co-first author: 3; Corresponding author and co-corresponding author: 2; Zone 3 / Zone 4: First author and co-first author: 2; Corresponding author: 1; Zone 5 / Zone 6: First author and co-first author: 1; Corresponding author: 1.

[0147] Special Cases: 1. If an article is estimated to be of low quality but is actually of high quality, the first author and the corresponding author will each receive an additional half of the extra points awarded, and the remaining 50% will be divided among the participating centers; 2. If an article is estimated to be of high quality but is actually of low quality, the first author and the corresponding author will each receive half of the difference in points deducted, and the remaining 50% will be divided among the participating centers; 3. The first clinical research project entered by the same patient data is given 100% points, the second clinical research project is given 50% points, and the third clinical research project and subsequent ones are given 25% points; 4. If a second article is published for the same clinical research project, the score for each patient will be calculated based on the journal score of the second article, and the score for each center will be given based on the number of patients enrolled (the first author and the corresponding author will be deducted first).

[0148] 5. The initiator of the clinical research has the priority to choose the first author and the corresponding author.

[0149] 6. The articles purchased with points must be related to the clinical research data entered, and the clinical research data not entered cannot be purchased with points.

[0150] After the completion of the clinical research project, the research results can be uploaded to the platform (research achievement interface) for review, and the corresponding points will be deducted from the user account. The review of the research results includes two stages: preliminary review and final review.

[0151] In the preliminary review stage, the medical group leader uploads the article type (estimated level), title, and associates the corresponding clinical research project, and fills in the author information.

[0152] In the review stage, the medical group leader uploads the proof materials of the research results, such as article links, screenshots, etc. The review stage will review the article type, etc. If the level of the relevant article is higher than the estimated publication article level, the integral is rewarded according to the reward and punishment mechanism. If the level of the relevant article is lower than the preset publication article level, the integral is deducted according to the reward and punishment mechanism.

[0153] For example, Zhang San as a medical group leader fills in the research plan in the research filling stage, and fills in A magazine of one area for submission. The platform calculates that one patient enters the group to get 100 points, the sample size of the research is 200, and the total integral of the research is 20000. After the research results are obtained, Zhang San's research can only reach B magazine of two areas. The platform calculates that one patient enters the group to get 50 points, the sample size of the research is 200, and the total integral of the research is 10000. The difference is 10000 points. The research person in charge is usually the first author or co-first author of the article. If the article has a first author and a co-first author, Zhang San should deduct (10000*50%) / 2=2500 points.

[0154] Through the integral module of the platform, the actual contribution of each center in the clinical research project can be quantified, and the contribution degree of the final results (such as published articles) of the project can be provided with strong evidence support, which avoids the subjective error introduced by artificially determining the contribution of each participant, provides a quantifiable and traceable environment for the participants of the clinical research project, so that each researcher can have the possibility to carry out (initiate or participate in) the clinical research project. The originally idle patient data is turned into treasure and is included in the clinical research project to make contributions to the project research, and the work load can be reflected in each participating project according to the actual labor, which can encourage researchers to participate in the clinical research project to a certain extent, and improve the efficiency of the clinical research project. In addition, the integral acquisition rule can constrain the medical group leader to more strictly evaluate the advanced degree of the project, thereby indirectly assisting the clinical research project to achieve the ideal results, realize the value of the clinical research, and improve the objectivity of the results of the clinical research project.

[0155] According to the idea of the application, the embodiment of the application further provides a computer program product, which comprises a computer program, and the computer program is run by a processor to run the multi-center clinical research data management platform.

[0156] The application is not limited to the foregoing specific embodiments. The application extends to any new feature disclosed in the specification or any new combination, as well as any new method or process step disclosed or any new combination.

Claims

1. A multi-center clinical research data management platform, characterized by, The application relates to a clinical research management system and method. The application comprises: a registration module configured to receive a user registration request, create a user account, the user including a main center as a medical group leader and a sub-center as a non-medical group leader; a benchmark library module configured to construct and manage a benchmark library; wherein the work of constructing the benchmark library comprises: constructing at least one CRF form under each benchmark library respectively; the work of managing the benchmark library comprises: selecting multiple basic components for combination for each CRF form in a pre-constructed basic component set, and entering patient data; a clinical research module configured to initiate and manage a clinical research project; wherein initiating the clinical research project comprises newly creating a clinical research project for the main center and adding the sub-center to the newly created clinical research project; managing the clinical research project comprises receiving newly added patient data of the main center and the sub-center, and analyzing the patient data of the clinical research project; 2. The multi-center clinical research data management platform of claim 1, wherein, a credit module configured to construct a credit rule and manage the credit of each user in the clinical research project according to the credit rule.

3. The multi-center clinical research data management platform of claim 2, wherein, After the patient data is entered, the benchmark library module matches whether the patient data meets the clinical research project that the user is currently participating in through a pre-configured pre-entry method to generate a pre-entry clinical research list. The method for matching whether the patient data meets the clinical research project that the user is currently participating in through the pre-configured pre-entry method comprises: adopting a four-layer intelligent matching architecture to match the patient data with the enrollment criteria set for the clinical research project; wherein the four-layer intelligent matching architecture comprises a data preprocessing layer, a feature calculation layer, an intelligent matching layer and a result optimization layer; the data preprocessing layer uniformly standardizes the patient data of participants of different clinical research projects; the feature calculation layer extracts five-dimensional features of the patient data from a demographic dimension, a clinical pathology dimension, a molecular typing dimension, a treatment response dimension and a time sequence dynamic dimension, each of which comprises multiple sub-features; the intelligent matching layer matches the patient data and the enrollment criteria by adopting a multi-granularity attention matching mechanism, a medical ontology enhanced semantic matching mechanism and a time sequence dynamic feature matching mechanism; 4. The multi-center clinical research data management platform of claim 3, wherein, the result optimization layer takes the matched pre-entry clinical research list as a reference to feed back the feature weight in the feature extraction stage and the threshold in the feature matching process. The method for matching the patient data and the enrollment criteria by adopting the multi-granularity attention matching mechanism comprises: feature-level matching: different methods are adopted according to the types of the sub-features: for numerical features, an adaptive matching algorithm is adopted for matching; for classification features, a medical ontology is used to calculate a semantic similarity for matching; for time sequence features, a dynamic time warping algorithm is used for matching; dimension-level matching: the importance weight of each dimension feature is automatically learned by applying an attention mechanism considering the clinical correlation constraint between the dimension features; patient-level matching: priori knowledge of clinical experts is introduced to adjust the weight distribution; the matching degrees of the dimension features are weighted and fused according to the obtained maximum weight of each dimension feature to obtain the matching degree under the multi-granularity attention matching mechanism. According to the matching degree of patient data from high to low, the patient data with matching degree not reaching the matching degree threshold is generated into a candidate patient list.

5. The multi-center clinical research data management platform of claim 4, wherein, In the feature level matching, the feature weight templates for different types of clinical research projects are pre-constructed, and for the matching degree of each feature, the weight indicated by the feature weight template corresponding to the type thereof is used for weighting.

6. The multi-center clinical research data management platform of claim 3, wherein, The method for matching patient data and enrollment criteria by using a medical ontology enhanced semantic matching mechanism comprises: Breast disease ontology construction: a hierarchical ontology structure containing diagnosis, treatment, pathology and gene concepts is established; and semantic relationships between concepts are defined; Semantic similarity calculation: diagnosis term similarity is calculated based on ontology path distance, wherein the ontology path distance is calculated by weighting based on concept information content and clinical relevance in the diagnosis term; patient data with similarity not reaching a set threshold is excluded; Treatment path matching: logical consistency between patient data and clinical research projects is analyzed; patient data with logical consistency not reaching a set threshold is excluded.

7. The multi-center clinical research data management platform of claim 3, wherein, The method for matching patient data and enrollment criteria by using a time sequence dynamic feature matching mechanism comprises: Time sequence analysis: key time point features of disease progression are extracted; the consistency of key time point biomarker change trends with clinical research projects is analyzed; patient data with consistency not reaching a set threshold is excluded; Dynamic inclusion and exclusion criteria support degree analysis: the support degree of time-related inclusion and exclusion criteria of clinical research projects is calculated; patient data with support degree not reaching a set threshold is excluded.

8. The multi-center clinical research data management platform of claim 1, wherein, When initiating or managing a clinical research project, the clinical research module gives the user establishing the clinical research project the permission of project configuration and CRF form management, wherein: The project configuration comprises configuration of research information, center and personnel configuration, demographic information, automatic coding, automatic follow-up and randomization; The CRF form management comprises adding, importing, exporting and deleting CRF forms required by the research, wherein the added CRF form is combined by adding multiple basic components from the basic component set.

9. The multi-center clinical study data management platform of claim 8, wherein, The configuration of randomization comprises: Setting grouping information and sample size; According to the needs of different clinical research projects, patient data is randomly allocated to different research groups, which comprises: Grouping configuration: blind selection, including non-blind, single-blind or double-blind; Randomization scheme configuration: random methods include simple random, block random and stratified block random; wherein, the block random needs to configure block length, and the stratified block random needs to configure block length, stratified mechanism and stratified factor; Random number configuration: prefix, project serial number and suffix are configured.

10. A computer program product, characterised in that, The computer program is run by a processor, and the multi-center clinical research data management platform of any one of claims 1-9 is run. The computer program is run by a processor, and the multi-center clinical research data management platform of any one of claims 1-9 is run.

Images