A diagnosis and treatment rule auditing method based on real-time confidence threshold judgment

Abstract

The application discloses a diagnosis and treatment rule auditing method based on real-time confidence threshold judgment, relates to the medical information technology field, and comprises the following steps: real-time monitoring of a triggering event and collection of diagnosis and treatment context data associated with the triggering event, encapsulation of the diagnosis and treatment context data into an auditing context package according to a unified time stamp and a source identifier; comparison of rule confidence scores through a dynamic confidence threshold set, screening of a target rule set, construction of a rule competition relationship structure, calculation of a competition weight according to a risk propagation intensity and a compliance dominance degree, execution of a cause-effect competition decision, and recording of a threshold mapping track to form traceable auditing records. The application constructs a confidence threshold model and performs double-domain nonlinear coupling mapping and interval projection calculation in a threshold candidate interval, so that the rule confidence scores and resource pressure constraint parameters form a dynamic response relationship in a unified capacity evaluation space.

Description

Technical Field

[0001] This invention relates to the field of medical information technology, and in particular to a method for auditing diagnosis and treatment rules based on real-time confidence threshold judgment. Background Technology

[0002] With the development of medical informatization and refined management, medical practices are gradually being standardized and controlled through electronic medical records, prescription management, and clinical pathways. The construction of medical practice rule bases and automated auditing technologies are constantly evolving. Existing technologies typically use rule matching, field validation, and risk scoring mechanisms to perform real-time or near-real-time compliance analysis of diagnostic records, medication practices, and test results. They also incorporate confidence level calculations and tiered early warning strategies to improve the automation level of auditing and the efficiency of risk identification.

[0003] In scenarios with concurrent triggering of multiple rules and limited audit resources, existing technologies often use fixed thresholds or single-variable linear adjustment methods for rule screening. They fail to model the evolution characteristics of rule confidence and audit capacity constraints in a unified manner, resulting in a lack of dynamic coupling mechanism between threshold adjustment and resource load. This makes it difficult to maintain a stable balance between audit trigger intensity and review capability in a complex competitive rule environment, thus forming the main technical problem of rule screening fluctuation and load imbalance. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a diagnostic and treatment rule auditing method based on real-time confidence threshold judgment to solve the problem that it is difficult to dynamically couple and adjust rule confidence and audit capacity constraints.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a method for auditing medical rules based on real-time confidence threshold judgment, comprising: real-time monitoring of triggering events and collection of medical context data associated with the triggering events, encapsulating them into an audit context package according to a unified timestamp and source identifier; performing auditable parsing and field mapping on the medical rule base based on the audit context package to generate a set of rule audit objects, extracting corresponding evidence fragments and missing marker sets, mapping them into rule evidence data, and performing multi-dimensional credible representation calculation on the set of rule audit objects to generate a rule confidence score; calculating resource pressure constraint parameters based on the review capacity and the number of rules to be audited in the current audit cycle, constructing a pressure confidence threshold model, and performing dual-domain coupling mapping and capacity projection on the resource pressure constraint parameters and rule confidence scores to generate a dynamic confidence threshold set; comparing the rule confidence scores through the dynamic confidence threshold set, filtering the target rule set, constructing a rule competition relationship structure, calculating the competition weight based on the risk propagation intensity and compliance dominance, executing causal competition adjudication, and simultaneously recording the threshold mapping trajectory to form a traceable audit record.

[0007] As a preferred embodiment of the diagnostic and treatment rule auditing method based on real-time confidence threshold judgment described in this invention, the triggering events include diagnostic data records, prescription generation records, test data feedback, and discharge settlement records.

[0008] As a preferred embodiment of the diagnostic and treatment rule auditing method based on real-time confidence threshold judgment described in this invention, the specific steps for forming the audit context package are as follows: Based on the triggering event, read the diagnosis and treatment context data, and construct audit data atoms according to the unified timestamp, source identifier, event type identifier, and time index identifier; The audit data atoms are associated and assembled based on the time index identifier and the source identifier to form an audit context package.

[0009] As a preferred embodiment of the diagnostic and treatment rule auditing method based on real-time confidence threshold judgment described in this invention, the specific steps for generating the rule audit object set are as follows: Based on the event type identifier and time index identifier, the candidate rule entries corresponding to the event type identifier are retrieved from the diagnosis and treatment rule base, and the trigger condition field, constraint condition field, exception clause field, evidence field list and action template field are parsed one by one, and the rule audit object is generated according to the unified field structure. Based on the evidence field list, the diagnosis and treatment context data in the audit context package are aligned with the evidence location and mapped, and written into the rule audit object to form a set of rule audit objects; The diagnostic and treatment rule base refers to a database that contains medical diagnostic and treatment standards and norms.

[0010] As a preferred embodiment of the diagnostic and treatment rule auditing method based on real-time confidence threshold judgment described in this invention, the specific steps for generating rule confidence scores are as follows: Based on the evidence field list, locate and extract the diagnosis and treatment context data from the audit context package to form a set of evidence fragments, and at the same time generate a set of missing markers for the evidence fields that were not matched; The set of evidence fragments and the set of missing markers are encapsulated in a unified structure and mapped into rule-based evidence data; Based on rule-based evidence data, the rule audit object set is subjected to multi-dimensional credible characterization calculation and fusion of evidence integrity, consistency, timeliness and source credibility to generate rule confidence score.

[0011] As a preferred embodiment of the diagnostic and treatment rule auditing method based on real-time confidence threshold judgment described in this invention, the specific steps for calculating the resource pressure constraint parameters are as follows: Read the review task processing records of each time slice in the current audit cycle and count the maximum number of reviewable rules and the actual number of reviews completed in each time slice to form a review capacity status sequence; Based on the set of rule audit objects, the number of newly added rules to be audited and the cumulative number of rules to be audited in each time slice within the current audit period are read, and the change range is calculated to form a rule load status sequence; Based on the verified capacity state sequence and the rule load state sequence, joint constraint mapping and capacity projection calculation are performed to obtain resource pressure constraint parameters.

[0012] As a preferred embodiment of the diagnostic and treatment rule auditing method based on real-time confidence threshold judgment described in this invention, the specific process of constructing the confidence threshold model is as follows: A hierarchical structure is constructed through an input mapping layer, a pressure-confidence coupling calculation layer, a capacity constraint determination layer, and a threshold output layer. The input mapping layer receives resource pressure constraint parameters and rule confidence scores and converts them into pressure-confidence state vectors. The pressure-trust coupling computation layer performs confidence evolution trend calculation and resource occupancy sensitivity mapping on the pressure-trust state vector, and forms threshold candidate intervals; The capacity constraint determination layer performs a capacity feasibility check and capacity projection correction on the threshold candidate interval, and passes the threshold parameters that meet the capacity constraint conditions to the threshold output layer for encapsulation and output, forming the pressure-sensitive threshold model.

[0013] As a preferred embodiment of the diagnostic and treatment rule auditing method based on real-time confidence threshold judgment described in this invention, the specific process of generating the dynamic confidence threshold set is as follows: The threshold candidate interval is read according to the pressure threshold model, and the threshold candidate interval, rule confidence score and resource pressure constraint parameter are aligned according to the rule identifier to form a threshold projection input tuple; Based on the threshold projection input tuple, a two-domain nonlinear coupling mapping is performed on the resource pressure constraint parameters and the rule confidence score, and interval projection calculation is performed in the threshold candidate interval to generate the projection threshold. Perform a capacity consistency check on the projection thresholds, perform convergence correction on projection thresholds that exceed capacity constraints, and encapsulate them according to rules, outputting a dynamic confidence threshold set.

[0014] As a preferred embodiment of the diagnostic and treatment rule auditing method based on real-time confidence threshold judgment described in this invention, the specific process of constructing the rule competition relationship structure is as follows: Based on the rule identifier, read the rule confidence score and projection threshold one by one, perform threshold comparison, mark the rule audit objects whose rule confidence scores are lower than the corresponding projection threshold, and write them into the target rule set; Based on the target rule set, the evidence field identifier, treatment behavior identifier, and time index identifier corresponding to each rule audit object are read. The rule audit objects that act on the same treatment behavior and share the same evidence field are matched to form rule association pairs. Based on the evidence source credibility identifier, confidence change trend identifier, and time sequence identifier of the rule association pair and the rule audit object, a directed mapping is performed to generate competitive relationship edges. According to the competitive relationship edges, the rule audit object is structured according to the rule identifier to form a rule competitive relationship structure.

[0015] As a preferred embodiment of the diagnostic and treatment rule auditing method based on real-time confidence threshold judgment described in this invention, the specific process of simultaneously recording the threshold mapping trajectory to form a traceable audit record is as follows: Based on the structure of the competitive relationship of the rules, the rule associations with competitive relationships are read one by one, and the risk propagation path of the rules is analyzed based on the credibility of the evidence source, the confidence change trend, and the time index of the diagnosis and treatment behavior, and risk propagation identifier is generated. Based on the set of rule audit objects, the constraint field and exception clause field are read, and the field matching and exception condition verification are performed in combination with the rule evidence data to generate constraint matching identifier and exception clause hit identifier. Based on the risk propagation identifier, constraint matching identifier, and exception clause hit identifier, a competition weight mapping is performed to generate a competition weight identifier. Based on the competition weight identifier, a causal competition ruling is performed on rules with competing relationships to generate a causal competition ruling correspondence. Based on the rule confidence score, projection threshold, and threshold candidate interval of the causal competition adjudication correspondence, the threshold mapping trajectory is recorded in time index order and written into the audit trajectory field to form a traceable audit record.

[0016] The beneficial effects of this invention are as follows: by constructing a pressure confidence threshold model and performing dual-domain nonlinear coupling mapping and interval projection calculation within the threshold candidate interval, the rule confidence score and resource pressure constraint parameters form a dynamic response relationship within a unified capacity evaluation space; by combining and mapping the confidence evolution trend with resource occupancy sensitivity, the adaptive adjustment of the rule-level threshold position is achieved while maintaining the threshold boundary constraints; and the controllable correspondence between the number of triggering rules and the review capability is ensured through capacity consistency verification and convergence correction mechanisms, thereby achieving smooth adjustment of audit intensity and stable operation of the load. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a diagnostic and treatment rule auditing method based on real-time confidence threshold judgment.

[0019] Figure 2 A flowchart generated for the rule audit object.

[0020] Figure 3 This is a flowchart for calculating resource pressure.

[0021] Figure 4 A flowchart for generating dynamic thresholds. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 This is one embodiment of the present invention, which provides a method for auditing diagnosis and treatment rules based on real-time confidence threshold judgment, including the following steps: S1: Real-time monitoring of triggered events and collection of diagnostic context data associated with the triggered events, encapsulating them into audit context packages according to a unified timestamp and source identifier.

[0026] S1.1: Based on the triggering event, read the diagnosis and treatment context data, and construct audit data atoms according to the unified timestamp, source identifier, event type identifier and time index identifier.

[0027] Specifically, based on the triggering event, the diagnosis and treatment context data associated with the triggering event is read, and a unified timestamp and source identifier are assigned to each data item to construct audit data atoms; the source identifier, event type identifier, and time index identifier corresponding to each audit data atom are read and written into the context field of each audit data atom.

[0028] It should be noted that triggering events include diagnostic data records, prescription generation records, laboratory data feedback, and discharge settlement records; treatment context data includes patient information, treatment records, medication and examination details, key indicators, and previous audit records.

[0029] S1.2: Associate and assemble the audit data atoms according to the time index identifier and the source identifier to form an audit context package.

[0030] Specifically, based on the time index identifier and the source identifier, the audit data atoms are matched with their corresponding time index identifier and source identifier. Each audit data atom is associated with the time sequence and source information. The associated audit data atoms are then assembled in sequence to form a complete audit context package.

[0031] S2: Based on the audit context package, the diagnosis and treatment rule base is parsed for auditability and field mapping, generating a set of rule audit objects, extracting the corresponding evidence fragments and missing marker sets, mapping them into rule evidence data, and performing multi-dimensional credible characterization calculation on the set of rule audit objects to generate rule confidence scores.

[0032] S2.1: Based on the event type identifier and time index identifier, retrieve the candidate rule entries corresponding to the event type identifier from the diagnosis and treatment rule base, and parse the trigger condition field, constraint condition field, exception clause field, evidence field list and action template field one by one, and generate rule audit objects according to a unified field structure.

[0033] Specifically, based on the event type identifier and time index identifier, the rule entries corresponding to the event type identifier and valid within the time period corresponding to the time index identifier are retrieved from the diagnosis and treatment rule base as candidate rule entries. The trigger condition field, constraint condition field, exception clause field, evidence field list, and action template field are read one by one for each candidate rule entry. The content of each field is processed by aligning field names, validating field types, and normalizing field positions. The trigger condition field, constraint condition field, exception clause field, evidence field list, and action template field are written into the same rule record according to a unified field structure, forming a rule audit object that corresponds one-to-one with the candidate rule entry.

[0034] The diagnosis and treatment rule base refers to a database containing medical diagnosis and treatment standards and norms; candidate rule entries include trigger condition fields, constraint condition fields, exception clause fields, evidence field list, and action template fields; medical diagnosis and treatment standards are set by organizing clinical diagnosis and treatment norms, diagnosis and treatment management systems, and current regulatory compliance requirements into entries according to a unified rule field structure and then writing them into the diagnosis and treatment rule base.

[0035] S2.2: Based on the evidence field list, perform field alignment and evidence location mapping on the diagnosis and treatment context data in the audit context package, and write it into the rule audit object to form a set of rule audit objects.

[0036] Specifically, based on the evidence field list, the diagnosis and treatment context data is matched one-to-one with the fields in the evidence field list, and the diagnosis and treatment data for each field is aligned with the evidence fields required by the rules; the aligned diagnosis and treatment data is written into the corresponding rule audit object to form a set of rule audit objects.

[0037] S2.3: Based on the evidence field list, locate and extract the diagnosis and treatment context data from the audit context package to form a set of evidence fragments, and generate a set of missing markers for the evidence fields that were not matched.

[0038] Specifically, based on the evidence field list, the medical context data is first retrieved item by item in the audit context package according to the field name, field type, and corresponding time index range. The retrieved content is then matched with fields. Only medical context data with consistent field identifiers and corresponding time positions are considered valid extraction content. Each valid extraction content is then assigned to the evidence fragment set according to the field order in the evidence field list. At the same time, evidence fields for which no corresponding medical context data was retrieved, whose field content is empty, or whose field type does not match are written to the missing marker set to indicate the missing field location and missing status.

[0039] S2.4: Encapsulate the set of evidence fragments and the set of missing markers into a unified structure and map them into rule-based evidence data.

[0040] Specifically, when encapsulating the evidence fragment set and the missing marker set according to a unified structure, the field order is first determined according to the evidence field list in the rule audit object. The evidence fragment or missing marker corresponding to each evidence field is written into the same field position, while retaining the corresponding field identifier, time index identifier, and source identifier to indicate the source and time sequence position of the evidence. Field-level mapping is performed on the encapsulated content to match each field position with the evidence field list in the rule audit object one by one, thus obtaining the rule evidence data.

[0041] S2.5: Based on rule-based evidence data, perform multi-dimensional credible characterization calculations and fusion on the rule audit object set to generate rule confidence scores, considering evidence integrity, consistency, timeliness, and source credibility.

[0042] Specifically, based on the rule-based evidence data, a multi-dimensional credible representation calculation is performed on the rule audit object set. The corresponding rule-based evidence data is read one by one according to the rule audit object in the rule audit object set. Based on the evidence field list, the total number of evidence fields that should be hit and the actual number of evidence fields that are hit are counted. The ratio of the actual number of evidence fields hit to the total number of evidence fields that should be hit is used as the evidence completeness. The value relationships, temporal relationships and correlations between different fields in the rule-based evidence data are compared item by item. The number of consistent items and the total number of items participating in the comparison are counted. The ratio of the number of consistent items to the total number of items participating in the comparison is used as the consistency.

[0043] The evidence time difference is calculated based on the time index identifier in the rule evidence data and the time position of the triggering event, and the evidence time difference is mapped to a timeliness characterization value. When the evidence time difference increases, the corresponding timeliness characterization value decreases. The source credibility is obtained by mapping the basic credibility level of the source type corresponding to the source identifier and the review accuracy rate in the historical credibility record.

[0044] The multi-dimensional credible representations corresponding to the integrity, consistency, timeliness, and source credibility of evidence are written into the corresponding rule audit objects in the rule audit object set, and the geometric fusion relationship is performed on the multi-dimensional credible representations to generate rule confidence scores.

[0045] Among them, the rule confidence score represents the reliability of the evidence supporting the rule audit object and the credibility of the rule judgment under the current triggering event and the current audit context.

[0046] S3: Calculate resource pressure constraint parameters based on the review capacity and the number of rules to be audited in the current audit cycle, construct a pressure confidence threshold model, and perform two-domain coupling mapping and capacity projection on the resource pressure constraint parameters and rule confidence scores to generate a dynamic confidence threshold set.

[0047] S3.1: Read the review task processing records of each time slice in the current audit period and count the maximum number of reviewable rules and the actual number of reviews completed in each time slice to form a review capacity status sequence.

[0048] Specifically, the time slice range is obtained according to the time division rules of the current audit cycle. Records that fall into the corresponding time slice are filtered one by one from the review task processing records according to the time slice range. The review task processing records after filtering are calculated separately for the time slice and the processing status of the review task.

[0049] The maximum number of rules that can be reviewed is calculated by converting the processing time within the time slice. The actual number of reviewed rules is obtained by counting each record whose processing status is displayed as completed within the time slice. The time index identifier, the maximum number of rules that can be reviewed, and the actual number of reviewed rules are arranged in chronological order for each time slice and written into the review capacity status sequence.

[0050] It should be noted that the review capacity status sequence refers to a set of time-series records arranged in time slice order, where each record includes a time index identifier, the maximum number of rules that can be reviewed, and the actual number of reviews completed.

[0051] Verifiable rules refer to rule audit objects that have entered the review task processing record within the corresponding time slice of the current audit cycle, have a valid rule identifier and processing status, and have not been revoked, exempted by exception, or repeatedly merged and removed. The rule audit objects selected are defined as verifiable rules by filtering and judging the time index identifier, processing status identifier, and task validity identifier in the review task processing record.

[0052] S3.2: Based on the set of rule audit objects, read the number of newly added rules to be audited and the cumulative number of rules to be audited in each time slice within the current audit period, and calculate the change range to form a rule load status sequence.

[0053] Specifically, based on the set of rule audit objects, the time slice range is obtained according to the time division rules of the current audit cycle. Rule audit objects that fall into the corresponding time slice and are in the pending audit state are filtered one by one from the set of rule audit objects according to the time slice range, and the number of newly added pending audit rules is counted. At the same time, the number of rules that are still in the pending audit state and have not been reviewed in the previous time slice is added to the number of newly added pending audit rules in the current time slice in chronological order to obtain the cumulative number of pending audit rules.

[0054] Perform difference calculation on the cumulative number of rules pending review for adjacent time slices to obtain the change range, and write the time index identifier, the number of newly added rules pending review, the cumulative number of rules pending review, and the change range corresponding to each time slice into the rule load status sequence in chronological order.

[0055] S3.3: Perform joint constraint mapping and capacity projection calculation based on the verification capacity state sequence and the rule load state sequence to obtain resource pressure constraint parameters.

[0056] Specifically, based on the time index identifier, the maximum number of reproducible rules and the actual number of completed reviews in the review capacity status sequence are aligned with the number of newly added rules pending review, the cumulative number of rules pending review, and the magnitude of change in the rule load status sequence on a time-slice basis. Joint constraint mapping is performed on the aligned time-slice records to map the processing capacity in the review capacity status sequence and the pending growth pressure in the rule load status sequence to the same capacity constraint evaluation space. Capacity projection calculation is performed on the capacity constraint values ​​of each mapped time slice. By comparing the deviation between the processing capacity and the pending growth pressure of each time slice, abnormal fluctuations are compressed and the boundary convergence is processed. The capacity constraint values ​​calculated by capacity projection of each time slice are summarized in chronological order to obtain the resource pressure constraint parameters.

[0057] It should be noted that the resource pressure constraint parameter refers to the constraint characterization quantity obtained by calculating the joint constraint mapping and capacity projection of the verification capacity state sequence and the rule load state sequence in the same capacity constraint evaluation space.

[0058] The processing capacity is obtained by reading the maximum number of reproducible rules and the actual number of reproducible rules corresponding to the same time slice in the reproducible capacity state sequence, and using these as the capacity representation of that time slice.

[0059] The pressure of pending rule growth is obtained by reading the number of newly added pending rules, the cumulative number of pending rules, and the magnitude of change in the same time slice in the rule load state sequence, and using these as load characteristics for that time slice.

[0060] The joint constraint mapping formula is: ; In the formula, Indicates the first Time slice capacity constraint value, Indicates the first The processing capability of time slices, Indicates the first The pressure of increasing censorship pressure for time-limited films, This represents the joint constraint mapping function.

[0061] The formula for calculating capacity projection is: ; In the formula, Indicates the first Projection capacity constraint value of time slice, Indicates the upper and lower boundaries of the capacity constraint. This represents the capacity projection function.

[0062] The resource pressure constraint parameter is given by the following formula: ; In the formula, Indicates the first Resource pressure constraints parameters for time slices, Indicates the total number of time slices.

[0063] Among them, the capacity constraint value, processing capacity, and pending growth pressure in the joint constraint mapping formula, capacity projection calculation formula, and resource pressure constraint parameter formula are all quantified values. The quantities processed all represent the rule processing capacity and rule load pressure characterization quantities in the same capacity constraint evaluation space, and the dimensions remain consistent before and after the calculation.

[0064] S3.4: A hierarchical structure is constructed through an input mapping layer, a pressure-confidence coupling calculation layer, a capacity constraint determination layer, and a threshold output layer. The input mapping layer receives resource pressure constraint parameters and rule confidence scores and converts them into pressure-confidence state vectors.

[0065] Specifically, the input mapping layer, pressure-confidence coupling calculation layer, capacity constraint determination layer, and threshold output layer are established in the order of "input mapping layer → pressure-confidence coupling calculation layer → capacity constraint determination layer → threshold output layer" to establish a forward-backward transmission relationship. The input mapping layer reads the confidence score of each rule according to the rule identifier and writes the resource pressure constraint parameter corresponding to the current audit period into the corresponding position of each rule. Field alignment, field type unification, and field order concatenation are performed on the resource pressure constraint parameter and rule confidence score. The pressure-confidence status vector is output according to the preset vector field structure.

[0066] The preset vector field structure is pre-defined by specifying the field order, field type, field length, normalization method, missing data filling method, and normalization value range for the resource pressure constraint parameter field and the rule confidence score field; the value range for the rule confidence score field is... The resource pressure constraint parameter field, after normalization before being written into the pressure state vector, has the following value range: .

[0067] S3.5: The pressure-trust coupling calculation layer performs confidence evolution trend calculation and resource occupancy sensitivity mapping on the pressure-trust state vector, and forms a threshold candidate interval.

[0068] Specifically, the resource pressure constraint parameter field and rule confidence score field in the pressure and confidence status vector are read one by one according to the rule identifier. The difference and direction determination of the rule confidence score field of the same rule identifier in adjacent time slices are performed according to the time index order to obtain the confidence change trend identifier. The pressure and confidence coupling calculation layer takes the resource pressure constraint parameter field corresponding to the same rule identifier as input, performs mapping processing on the change range and relative position of the resource pressure constraint parameter field in the current audit period, and performs interval position judgment and change range classification on the resource pressure constraint parameter field to distinguish whether the resource pressure constraint parameter field is in a low, medium or high occupancy state, and converts the resource pressure constraint parameter field into an occupancy sensitive identifier. The pressure and confidence coupling calculation layer combines the confidence change trend identifier and the occupancy sensitive identifier according to the rule identifier, and generates the lower threshold and upper threshold corresponding to the rule identifier according to the position relationship of the combined field, forming the threshold candidate interval.

[0069] S3.6: The capacity constraint determination layer performs a capacity feasibility check and capacity projection correction on the threshold candidate interval, and passes the threshold parameters that meet the capacity constraint conditions to the threshold output layer for encapsulation and output, forming the pressure-sensitive threshold model.

[0070] Specifically, the lower and upper bounds of the threshold in the threshold candidate interval are read one by one according to the rule identifier, and the feasibility of the review capacity is performed in combination with the resource pressure constraint parameters corresponding to the review capacity and the number of rules to be reviewed in the current audit period. In this process, the number of expected triggering rules is calculated based on the comparison relationship between the rule confidence score and the threshold parameters in the threshold candidate interval according to the rule identifier, and compared with the review capacity of the corresponding time slice in time slice order according to the time index.

[0071] For rule identifiers whose expected number of triggering rules exceeds the corresponding time slice review capacity, capacity projection correction is performed on the threshold candidate interval. While maintaining the order of the upper and lower bounds of the threshold candidate interval, the position of the threshold parameter in the threshold candidate interval is adjusted according to the rule identifier to reduce the expected number of triggering rules. The threshold parameter after passing the review capacity feasibility verification and capacity projection correction is passed to the threshold output layer according to the rule identifier and encapsulated and output according to a unified field structure. The pressure-information threshold model is formed by the input mapping layer, the pressure-information coupling calculation layer, the capacity constraint judgment layer and the threshold output layer.

[0072] The expected number of triggered rules is obtained by taking the current threshold parameter to be verified within the threshold candidate interval corresponding to each rule identifier, comparing the threshold parameter with the rule confidence score of each rule audit object in the same time slice, and counting the rule audit objects whose rule confidence score is lower than the threshold parameter.

[0073] S3.7: Read the threshold candidate interval according to the pressure confidence threshold model, and align the threshold candidate interval, rule confidence score and resource pressure constraint parameters according to the rule identifier to form a threshold projection input tuple.

[0074] Specifically, the threshold candidate interval is read according to the pressure confidence threshold model. The threshold parameters encapsulated in the threshold output layer of the pressure confidence threshold model are read one by one according to the rule identifier. The rule confidence score of the corresponding rule audit object is read according to the same rule identifier. The resource pressure constraint parameter corresponding to the current audit period is also read. Field alignment processing according to the rule identifier is performed on the threshold candidate interval, rule confidence score and resource pressure constraint parameter. The consistency of rule identifier, the unified field order and the unified field position are checked. The threshold candidate interval, rule confidence score and resource pressure constraint parameter corresponding to the same rule identifier are written into the same record to form a threshold projection input tuple.

[0075] S3.8: Based on the threshold projection input tuple, perform a two-domain nonlinear coupling mapping on the resource pressure constraint parameters and the rule confidence score, and perform interval projection calculation within the threshold candidate interval to generate the projection threshold.

[0076] Specifically, the threshold candidate interval, rule confidence score, and resource pressure constraint parameters are read one by one from the threshold projection input tuple according to the rule identifier. A two-domain nonlinear coupling mapping is performed on the resource pressure constraint parameters and the rule confidence score. The threshold offset direction and threshold offset magnitude are calculated based on the relative position of the resource pressure constraint parameters in the current audit period and the deviation position of the rule confidence score in the threshold candidate interval. Interval projection calculation is performed within the threshold candidate interval. The lower and upper bounds of the threshold candidate interval are used as boundary constraints to map the threshold offset direction and threshold offset magnitude to specific positions inside the threshold candidate interval. The threshold parameters corresponding to the mapped positions are written to the corresponding record of the rule identifier to generate the projected threshold.

[0077] Among them, the projection threshold refers to the rule-level dynamic threshold parameter obtained by mapping the resource pressure constraint parameters and the rule confidence score through a two-domain nonlinear coupling within the threshold candidate interval boundary constraints and writing it into the record corresponding to the rule identifier.

[0078] "Dual-domain nonlinear coupling mapping" refers to two domains: the resource pressure domain in which the resource pressure constraint parameters reside and the rule confidence domain in which the rule confidence score resides.

[0079] The threshold offset direction is determined based on the combination of resource pressure status and confidence deviation status, and one of three directions is selected: offset towards the upper bound of the threshold candidate interval, offset towards the lower bound of the threshold candidate interval, and maintain the current position.

[0080] The threshold offset magnitude is graded based on the degree of deviation of the combination of the relative position of the resource pressure constraint parameter and the deviation of the rule confidence score from the position, and one of the three magnitude levels of small offset, medium offset and large offset is selected.

[0081] A superior approach, compared to using a fixed threshold or performing univariate linear adjustment based solely on rule confidence scores, aligns the threshold candidate interval, rule confidence scores, and resource pressure constraint parameters according to rule identifiers by projecting the threshold input tuple, and performs dual-domain nonlinear coupling mapping and interval projection calculation to generate the projected threshold. This approach simultaneously responds to changes in rule confidence scores and resource pressure constraint parameters within the boundary constraints of the threshold candidate interval, improving the precision of threshold adjustment, rule-level adaptability, and audit load stability.

[0082] S3.9: Perform a capacity consistency check on the projection thresholds, perform convergence correction on projection thresholds that exceed capacity constraints, and encapsulate them according to the rules, outputting a dynamic confidence threshold set.

[0083] Specifically, the projection thresholds are read one by one according to the rule identifiers. The relationship between the review capacity and the number of rules to be reviewed in the current audit cycle is used to determine whether the number of rules triggered after the projection threshold participates in the threshold comparison falls within the capacity constraint range. The expected number of rules triggered by the projection thresholds within each time slice is counted in time index order and compared with the review capacity of the corresponding time slice. For projection thresholds where the expected number of triggered rules is higher than the review capacity of the corresponding time slice, convergence correction is performed. While maintaining the threshold sorting relationship corresponding to the rule identifiers and the boundary constraint relationship of the threshold candidate interval, the position of the projection thresholds in the threshold candidate interval is adjusted to reduce the expected number of triggered rules until it falls within the capacity constraint range. The projection thresholds, after passing the review capacity consistency check and convergence correction, are encapsulated into fields according to the rule identifiers and output as a dynamic confidence threshold set according to a unified field structure.

[0084] The dynamic confidence threshold set refers to the set of threshold parameters organized according to the rules. Each item in the dynamic confidence threshold set corresponds to the projection threshold of a rule audit object, and the projection thresholds in the dynamic confidence threshold set have been reviewed for capacity consistency verification and convergence correction.

[0085] S4: By comparing the confidence scores of rules with a dynamic confidence threshold set, the target rule set is selected, and a rule competition relationship structure is constructed. The competition weight is calculated based on the risk propagation intensity and compliance dominance to execute the causal competition adjudication. At the same time, the threshold mapping trajectory is recorded to form a traceable audit record.

[0086] S4.1: Read the confidence score and projection threshold of each rule according to the rule identifier, perform threshold comparison, mark the rule audit objects whose rule confidence score is lower than the corresponding projection threshold, and write them into the target rule set.

[0087] Specifically, based on the rule identifier, the rule confidence score in the rule audit object set is read one by one, and the projection threshold corresponding to the same rule identifier is read from the dynamic confidence threshold set. A one-to-one threshold comparison process is performed on the rule confidence score and the projection threshold to check the consistency of rule identifiers, the uniformity of comparison order, and the comparison relationship of each rule. When the rule confidence score is lower than the corresponding projection threshold, the rule audit object with the corresponding rule identifier is marked as an audit triggered state, and the rule identifier, rule confidence score, and projection threshold are written to the corresponding record position in the target rule set. When the rule confidence score is not lower than the corresponding projection threshold, the operation of writing to the target rule set is not performed, thus forming the target rule set.

[0088] S4.2: Based on the target rule set, read the evidence field identifier, treatment behavior identifier and time index identifier corresponding to each rule audit object, perform association matching on rule audit objects that act on the same treatment behavior and share evidence fields, and form rule association pairs.

[0089] Specifically, when reading the evidence field identifier, treatment behavior identifier, and time index identifier corresponding to each rule audit object based on the target rule set, the rule audit object records in the target rule set are extracted one by one according to the rule identifier. The rule audit objects in the target rule set are paired and matched using the treatment behavior identifier as the first matching condition and the evidence field identifier as the second matching condition. The matching conditions are: whether the treatment behavior identifier is consistent, whether the evidence field identifier has an intersection, and whether the time index identifier falls within the associatable time range. For rule audit objects that meet the conditions of "acting on the same treatment behavior and sharing evidence fields", the corresponding two rule identifiers and the associative basis are written into the same associative record to form a rule association pair.

[0090] It should be noted that the basis for association refers to the judgment information that proves that two rule audit objects can be formed into the same rule association pair.

[0091] The associative time range can be obtained by reading the time index identifier corresponding to the rule audit object and combining it with the time window associated with the diagnosis and treatment behavior.

[0092] S4.3: Based on the evidence source credibility identifier, confidence change trend identifier, and time sequence identifier of the rule association pair and the rule audit object, perform directed mapping to generate competitive relationship edges. Based on the competitive relationship edges, organize the rule audit object according to the rule identifier to form a rule competitive relationship structure.

[0093] Specifically, the system reads the two rule identifiers in each rule association pair one by one, and reads the source identifier, rule confidence score and time index identifier in the rule evidence data corresponding to the two rule identifiers respectively. Based on the source identifier, the system performs one-to-one matching in the pre-configured source credibility mapping rules, writes the credibility level corresponding to the source identifier into the evidence source credibility identifier, and generates a time sequence identifier based on the time index identifier sequence relationship.

[0094] The system compares the credibility of evidence sources of two rule audit objects, the direction and strength of the change in confidence trends of the two rule audit objects, and the chronological order of the two rule audit objects. Based on the comparison, it obtains the pointing relationship in the rule association pair and writes the two rule identifiers and the directed mapping basis into the competition relationship edge. The system summarizes the rule audit objects in the target rule set according to the rule identifier and reads the competition relationship edge associated with the rule identifier. The competition relationship edge is classified and sorted according to the rule identifier. The competition relationship edge record associated with each rule identifier is written into the same structure record to form the rule competition relationship structure.

[0095] It should be noted that the rule-based competition relationship structure refers to a directed relationship record structure organized according to rule identifiers, in which the competition relationship edge record associated with each rule identifier, as well as the pointing relationship and directed mapping basis in the competition relationship edge, are classified and written into the same structure record.

[0096] S4.4: Read the rule association pairs with competing relationships one by one according to the rule competition relationship structure, and analyze the rule risk propagation path based on the evidence source credibility identifier, confidence change trend identifier and diagnosis and treatment behavior time index identifier, and generate risk propagation identifier.

[0097] Specifically, based on the rule identifier, the competitive relationship edge records and directed mapping basis records corresponding to the rule association pairs are extracted from the rule competition relationship structure. The evidence source credibility identifier, confidence level change trend identifier, and treatment behavior time index identifier of the corresponding rule audit object in the target rule set are read according to the two rule identifiers in the rule association pair. The rule risk propagation path is analyzed based on the evidence source credibility identifier, confidence level change trend identifier, and treatment behavior time index identifier. The evidence source credibility identifier, the direction and strength of change corresponding to the confidence level change trend identifier, and the order of occurrence corresponding to the treatment behavior time index identifier are compared according to the two rule identifiers. Based on the comparison, the propagation direction and propagation priority relationship of the rule risk propagation path are obtained. The two rule identifiers corresponding to the rule association pair, the propagation direction and propagation priority relationship of the rule risk propagation path are written into the same record according to a unified field position to generate a risk propagation identifier.

[0098] It should be noted that the rule risk propagation path refers to the directed association path in which risk is transmitted from one rule audit object to another, determined based on the credibility identifier of the evidence source, the confidence trend identifier, and the time index identifier of the diagnosis and treatment behavior, among rule association pairs with competing relationships.

[0099] S4.5: Based on the set of rule audit objects, read the constraint field and exception clause field, combine the rule evidence data to perform field matching and exception condition verification, and generate constraint matching identifier and exception clause hit identifier.

[0100] Specifically, the constraint and exception fields in the rule audit object set are extracted one by one. The evidence field records in the rule evidence data corresponding to the same rule identifier are read, and field matching processing is performed on the constraint fields and rule evidence data. The evidence fields in the rule evidence data are compared item by item to see if they meet the value conditions, time conditions, and association conditions corresponding to the constraint fields. Exception condition verification is performed on the exception field and rule evidence data. The corresponding evidence fields in the rule evidence data are checked item by item to see if they meet the exception trigger conditions in the exception field. The matching status corresponding to the field matching processing is written to the constraint matching identifier, and the hit status corresponding to the exception condition verification is written to the exception clause hit identifier.

[0101] It should be noted that if the value condition, time condition, and associated condition corresponding to the constraint condition field are met, the constraint matching indicator will be set to a matched state; if any of the value condition, time condition, or associated condition corresponding to the constraint condition field is not met, the constraint matching indicator will be set to a mismatch state; if the exception trigger condition in the exception clause field is met, the exception clause hit indicator will be set to a hit state; if the exception trigger condition in the exception clause field is not met, the exception clause hit indicator will be set to a miss state.

[0102] S4.6: Perform competition weight mapping based on risk propagation identifier, constraint matching identifier and exception clause hit identifier, generate competition weight identifier, and perform causal competition adjudication on rules with competing relationships based on competition weight identifier, generating causal competition adjudication correspondence.

[0103] Specifically, when performing competitive weight mapping based on risk propagation identifier, constraint matching identifier, and exception clause hit identifier, for each set of rules with competitive relationships, the corresponding risk propagation identifier, constraint matching identifier, and exception clause hit identifier are read and jointly determined: the propagation direction and propagation priority relationship between rules are obtained based on the risk propagation identifier; the constraint matching identifier is used to determine whether the rule constraint is valid and participates in the adjudication; the exception clause hit identifier is used to determine whether the rule has entered the exception processing path; and the priority status after joint determination is written into the competitive weight identifier.

[0104] When executing causal competition adjudication based on competition weight identifiers, the two rules in the same rule association pair are assigned roles according to the priority corresponding to the competition weight identifiers. The rule that is retained first is associated with the rule that is postponed or suppressed, and the correspondence is written as the causal competition adjudication correspondence.

[0105] A better approach, compared to using preset priorities, fixed coverage order, or single risk level ranking to adjudicate conflicts between competing rules, is to perform competition weight mapping through risk propagation identifiers, constraint matching identifiers, and exception clause hit identifiers, and to perform causal competition adjudication on competing rules based on the competition weight identifiers. This approach can simultaneously reflect the impact of risk propagation paths, constraint matching status, and exception clause hit status on rule conflict adjudication, thereby improving the adaptability to complex scenarios of causal competition adjudication correspondence.

[0106] S4.7: Based on the rule confidence score, projection threshold and threshold candidate interval of the causal competition adjudication correspondence, record the threshold mapping trajectory in time index order and write it into the audit trajectory field to form a traceable audit record.

[0107] Specifically, the system reads the rule identifiers and causal competition adjudication correspondences of competing rules one by one, and reads the rule confidence score, projection threshold, threshold candidate interval, and time index identifier according to the same rule identifier. When recording the threshold mapping trajectory, the system first compares the relative positions of the projection threshold with the lower and upper bounds of the threshold candidate interval to obtain the positional relationship of the projection threshold within the threshold candidate interval. The system writes the rule identifier, time index identifier, rule confidence score, projection threshold, threshold candidate interval, and causal competition adjudication correspondence into the audit trajectory field according to the unified field position, and arranges the multiple audit trajectory field records corresponding to the same rule identifier in the order of time index to form a traceable audit record.

[0108] Among them, traceable audit records refer to a set of audit trajectory records organized in the order of rule identifiers and time indexes. Each audit trajectory record includes at least a rule identifier, a time index identifier, a rule confidence score, a projection threshold, a threshold candidate interval, and a causal competition adjudication correspondence.

[0109] In summary, this invention achieves a dynamic response relationship between rule confidence scores and resource pressure constraint parameters within a unified capacity evaluation space by: constructing a pressure threshold model and performing dual-domain nonlinear coupling mapping and interval projection calculation within the threshold candidate interval; by combining and mapping the confidence evolution trend with resource occupancy sensitivity, adaptive adjustment of rule-level threshold positions is achieved while maintaining threshold boundary constraints; and by using capacity consistency verification and convergence correction mechanisms, a controllable correspondence between the number of triggering rules and the review capability is ensured, thereby achieving smooth adjustment of audit intensity and stable operation of the load.

[0110] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for auditing treatment rules based on real-time confidence threshold judgment, characterized in that, include: Real-time monitoring of triggered events and collection of diagnostic context data associated with the triggered events, encapsulating them into audit context packages with a unified timestamp and source identifier; Based on the audit context package, the diagnosis and treatment rule base is parsed for auditability and field mapping, generating a set of rule audit objects, extracting the corresponding evidence fragments and missing marker sets, mapping them into rule evidence data, and performing multi-dimensional credible characterization calculation on the set of rule audit objects to generate rule confidence scores; Based on the review capacity and the number of rules to be audited in the current audit cycle, resource pressure constraint parameters are calculated, a pressure confidence threshold model is constructed, and a two-domain coupling mapping and capacity projection are performed on the resource pressure constraint parameters and the rule confidence scores to generate a dynamic confidence threshold set. By comparing the confidence scores of rules with a dynamic confidence threshold set, a set of target rules is selected, and a rule competition relationship structure is constructed. The competition weight is calculated based on the risk propagation intensity and compliance dominance to execute causal competition adjudication. At the same time, the threshold mapping trajectory is recorded to form a traceable audit record.

2. The method for auditing treatment rules based on real-time confidence threshold judgment as described in claim 1, characterized in that, The triggering events include diagnostic data recording, prescription generation records, test data feedback, and discharge settlement records.

3. The method for auditing treatment rules based on real-time confidence threshold judgment as described in claim 2, characterized in that, The specific steps for forming the audit context package are as follows: Based on the triggering event, read the diagnosis and treatment context data, and construct audit data atoms according to the unified timestamp, source identifier, event type identifier, and time index identifier; The audit data atoms are associated and assembled based on the time index identifier and the source identifier to form an audit context package.

4. The method for auditing treatment rules based on real-time confidence threshold judgment as described in claim 1, characterized in that, The specific steps for generating the set of auditable objects according to the rules are as follows: Based on the event type identifier and time index identifier, the candidate rule entries corresponding to the event type identifier are retrieved from the diagnosis and treatment rule base, and the trigger condition field, constraint condition field, exception clause field, evidence field list and action template field are parsed one by one, and the rule audit object is generated according to the unified field structure. Based on the evidence field list, the diagnosis and treatment context data in the audit context package are aligned with the evidence location and mapped, and written into the rule audit object to form a set of rule audit objects; The diagnostic and treatment rule base refers to a database that contains medical diagnostic and treatment standards and norms.

5. The method for auditing treatment rules based on real-time confidence threshold judgment as described in claim 4, characterized in that, The specific steps for generating the rule confidence score are as follows: Based on the evidence field list, locate and extract the diagnosis and treatment context data from the audit context package to form a set of evidence fragments, and at the same time generate a set of missing markers for the evidence fields that were not matched; The set of evidence fragments and the set of missing markers are encapsulated in a unified structure and mapped into rule-based evidence data; Based on rule-based evidence data, the rule audit object set is subjected to multi-dimensional credible characterization calculation and fusion of evidence integrity, consistency, timeliness and source credibility to generate rule confidence score.

6. The method for auditing treatment rules based on real-time confidence threshold judgment as described in claim 1, characterized in that, The specific steps for calculating the resource pressure constraint parameters are as follows: Read the review task processing records of each time slice in the current audit cycle and count the maximum number of reviewable rules and the actual number of reviews completed in each time slice to form a review capacity status sequence; Based on the set of rule audit objects, the number of newly added rules to be audited and the cumulative number of rules to be audited in each time slice within the current audit period are read, and the change range is calculated to form a rule load status sequence; Based on the verified capacity state sequence and the rule load state sequence, joint constraint mapping and capacity projection calculation are performed to obtain resource pressure constraint parameters.

7. The method for auditing treatment rules based on real-time confidence threshold judgment as described in claim 1, characterized in that, The specific process for constructing the pressure threshold model is as follows: A hierarchical structure is constructed through an input mapping layer, a pressure-confidence coupling calculation layer, a capacity constraint determination layer, and a threshold output layer. The input mapping layer receives resource pressure constraint parameters and rule confidence scores and converts them into pressure-confidence state vectors. The pressure-trust coupling computation layer performs confidence evolution trend calculation and resource occupancy sensitivity mapping on the pressure-trust state vector, and forms threshold candidate intervals; The capacity constraint determination layer performs a capacity feasibility check and capacity projection correction on the threshold candidate interval, and passes the threshold parameters that meet the capacity constraint conditions to the threshold output layer for encapsulation and output, forming the pressure-sensitive threshold model.

8. The method for auditing treatment rules based on real-time confidence threshold judgment as described in claim 7, characterized in that, The specific process for generating the dynamic confidence threshold set is as follows: The threshold candidate interval is read according to the pressure threshold model, and the threshold candidate interval, rule confidence score and resource pressure constraint parameter are aligned according to the rule identifier to form a threshold projection input tuple; Based on the threshold projection input tuple, a two-domain nonlinear coupling mapping is performed on the resource pressure constraint parameters and the rule confidence score, and interval projection calculation is performed in the threshold candidate interval to generate the projection threshold. Perform a capacity consistency check on the projection thresholds, perform convergence correction on projection thresholds that exceed capacity constraints, and encapsulate them according to rules, outputting a dynamic confidence threshold set.

9. The method for auditing treatment rules based on real-time confidence threshold judgment as described in claim 1, characterized in that, The specific process for constructing the rule-based competition relationship structure is as follows: Based on the rule identifier, read the rule confidence score and projection threshold one by one, perform threshold comparison, mark the rule audit objects whose rule confidence scores are lower than the corresponding projection threshold, and write them into the target rule set; Based on the target rule set, the evidence field identifier, treatment behavior identifier, and time index identifier corresponding to each rule audit object are read. The rule audit objects that act on the same treatment behavior and share the same evidence field are matched to form rule association pairs. Based on the evidence source credibility identifier, confidence change trend identifier, and time sequence identifier of the rule association pair and the rule audit object, a directed mapping is performed to generate competitive relationship edges. According to the competitive relationship edges, the rule audit object is structured according to the rule identifier to form a rule competitive relationship structure.

10. The method for auditing treatment rules based on real-time confidence threshold judgment as described in claim 1, characterized in that, The process of simultaneously recording the threshold mapping trajectory to form a traceable audit record is as follows: Based on the structure of the competitive relationship of the rules, the rule associations with competitive relationships are read one by one, and the risk propagation path of the rules is analyzed based on the credibility of the evidence source, the confidence change trend, and the time index of the diagnosis and treatment behavior, and risk propagation identifier is generated. Based on the set of rule audit objects, the constraint field and exception clause field are read, and the field matching and exception condition verification are performed in combination with the rule evidence data to generate constraint matching identifier and exception clause hit identifier. Based on the risk propagation identifier, constraint matching identifier, and exception clause hit identifier, a competition weight mapping is performed to generate a competition weight identifier. Based on the competition weight identifier, a causal competition ruling is performed on rules with competing relationships to generate a causal competition ruling correspondence. Based on the rule confidence score, projection threshold, and threshold candidate interval of the causal competition adjudication correspondence, the threshold mapping trajectory is recorded in time index order and written into the audit trajectory field to form a traceable audit record.

Images