Medical quality monitoring method and system based on artificial intelligence

Abstract

The invention discloses a medical quality monitoring method and system based on artificial intelligence, particularly relates to the technical field of artificial intelligence assisted medical decision, and is used for solving the problem that the existing static early warning threshold cannot collaboratively optimize the adaptation of a conflict target and a dynamic resource. The method comprises the steps of obtaining a monitoring target set and a resource constraint set in real time; analyzing real-time cascade influence intensity and resource linkage fluctuation values between targets with medical relevance; generating a target conflict incidence matrix and identifying a strong conflict coupling target group; activating a high conflict-low resource pressure state when the cascade influence intensity is greater than a threshold value, the linkage fluctuation value is greater than a fluctuation risk threshold value and a strong conflict target group exists; under the state, differential adjustment coefficients for different conflict targets are dynamically calculated based on the conflict sensitive factor and the resource margin factor so as to correct an early warning threshold and trigger risk early warning, dynamic balance of conflict target optimization and resource constraint is realized, and the early warning sensitivity and the false alarm suppression capability are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence-assisted medical decision-making technology, and more specifically, to a medical quality monitoring method and system based on artificial intelligence. Background Technology

[0002] In existing medical quality monitoring systems, artificial intelligence technology is used to analyze clinical data and generate risk warning signals. Warnings are usually triggered based on preset static rules or single-target models (such as optimizing only sensitivity or specificity). The warning thresholds are fixed and deployed after training with historical data. Medical institutions rely on such systems to identify risk events such as medication errors and postoperative complications in order to achieve quality control goals.

[0003] In existing technologies, fixed threshold early warning mechanisms cannot coordinate and optimize conflicting monitoring targets and dynamic resource constraints. This leads to a dilemma for the system in real clinical scenarios: if priority is given to ensuring early warning sensitivity, the false alarm rate will increase significantly, causing alarm fatigue among medical staff and crowding out limited medical resources; if the focus is on reducing the false alarm rate, the risk of missing high-risk events will increase, delaying the timing of clinical intervention. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a medical quality monitoring method and system based on artificial intelligence to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] An artificial intelligence-based medical quality monitoring method includes:

[0007] S1. Real-time acquisition of the monitoring target set and resource constraint set of the medical quality monitoring system;

[0008] S2. Analyze the real-time cascading impact intensity of any two targets in the monitoring target set, and analyze the chain fluctuation value of a single resource occupancy status change in the resource constraint set on the overall resource system.

[0009] S3. Identify the conflict correlation strength between different monitoring targets in the monitoring target set and generate a target conflict correlation matrix;

[0010] S4. When there are target pairs in the target conflict correlation matrix whose conflict correlation strength is greater than the correlation strength threshold, they are marked as strongly conflicting coupled target groups.

[0011] S5. When the real-time cascading impact intensity is greater than the cascading intensity threshold, the chain fluctuation value is greater than the fluctuation risk threshold, and there is a strongly conflicting coupled target group, it is identified as a high conflict-low resource pressure state.

[0012] S6. Under the condition of high conflict and low resource pressure, generate a dynamic adjustment coefficient for the early warning threshold based on the preset conflict sensitivity factor and resource adequacy factor to correct the current early warning threshold and trigger a medical risk warning.

[0013] Furthermore, the monitoring target set and resource constraint set of the medical quality monitoring system are acquired in real time, including:

[0014] Extract a list of preset monitoring target types from the medical quality monitoring system database. The list of monitoring target types includes sensitivity optimization targets and false alarm rate suppression targets.

[0015] The real-time parameter values ​​of the sensitivity optimization target and the false alarm rate suppression target are obtained to form a set of monitoring targets;

[0016] The resource identifier list is retrieved from the medical resource management platform. The resource identifier list includes consumable resource identifiers and occupied resource identifiers.

[0017] The real-time remaining available quantity corresponding to the resource consumption identifier and the real-time idle available capacity corresponding to the resource occupancy identifier are used to form a resource constraint set.

[0018] Furthermore, the analysis examines the real-time cascading impact strength of any two targets in the monitored target set, and analyzes the cascading fluctuations of a single resource occupancy change in the resource constraint set on the overall resource system, including:

[0019] For the first and second targets in the set of monitored targets, monitor the transmission rate and direction of the transmission of the real-time parameter value change of the first target to the real-time parameter value change of the second target, and generate the real-time cascading effect intensity based on the transmission rate and direction;

[0020] For the target resource node in the resource constraint set, identify the set of associated resource nodes that have a dependency relationship with the target resource node, and calculate the average decay of the real-time idle and available capacity of each resource node in the set of associated resource nodes when the occupancy status of the target resource node changes, as the cascading fluctuation value.

[0021] Furthermore, the intensity of real-time cascading effects is the vector product of the conduction rate and the conduction direction.

[0022] Furthermore, the conflict correlation strength between different monitoring targets in the monitoring target set is identified, and a target conflict correlation matrix is ​​generated, including:

[0023] For target pairs consisting of sensitivity optimization targets and false alarm rate suppression targets in the monitoring target set, extract the co-occurrence frequency of conflict alarm events between sensitivity optimization targets and false alarm rate suppression targets in historical data;

[0024] Calculate the absolute value of the difference between the real-time parameter value of the sensitivity optimization target and the real-time parameter value of the false alarm rate suppression target within the current time window;

[0025] The weighted sum of co-occurrence frequency and absolute value of difference is used as the conflict association strength;

[0026] Traverse all target pair combinations in the monitored target set and generate a conflict association strength matrix with the target pair as the row and column index as the target conflict association matrix.

[0027] Furthermore, when there are target pairs in the target conflict correlation matrix whose conflict correlation strength is greater than the correlation strength threshold, they are marked as strongly conflict-coupled target groups, including:

[0028] Traverse all target pair cells consisting of sensitivity optimization targets and false alarm rate suppression targets in the target conflict correlation matrix;

[0029] If the conflict correlation strength of the target pair cell is greater than the correlation strength threshold, the sensitivity optimization target identifier and false alarm rate suppression target identifier in the target pair cell are added to the conflict target identifier set;

[0030] Merge the target pair units that contain target identifiers with the same sensitivity optimization or the same false alarm rate suppression to generate a strongly conflict-coupled target group.

[0031] Furthermore, when the real-time cascading impact strength exceeds the cascading strength threshold, the chain fluctuation value exceeds the fluctuation risk threshold, and there is a strongly conflicting coupled target group, it is identified as a high-conflict-low-resource-pressure state, including:

[0032] The system compares the real-time cascading effect intensity value with the cascading intensity threshold in parallel, and also compares the chain fluctuation value with the fluctuation safety threshold.

[0033] Verify whether the strongly conflict-coupled target group contains sensitivity-optimized target identifiers and false alarm rate suppression target identifiers;

[0034] When the real-time cascading impact strength value is greater than the cascading strength threshold, the chain fluctuation value is greater than the fluctuation safety threshold, and there are sensitivity optimization target identifiers and false alarm rate suppression target identifiers in the strongly conflict-coupled target group, the high conflict-low resource pressure state identifier is activated.

[0035] Furthermore, under a high-conflict, low-resource-pressure state, a dynamic adjustment coefficient for the early warning threshold is generated based on preset conflict sensitivity factors and resource adequacy factors to correct the current early warning threshold and trigger a medical risk warning, including:

[0036] Obtain the scalar values ​​of the conflict sensitivity factor and resource margin factor corresponding to the activation of the high conflict-low resource pressure state identifier;

[0037] Take the natural logarithm of the conflict sensitivity factor scalar, and simultaneously calculate the square root of the resource margin factor scalar;

[0038] Multiply the natural logarithm result by the square root result to obtain the dynamic adjustment coefficient of the early warning threshold;

[0039] Multiply the current warning threshold by the warning threshold dynamic adjustment coefficient to obtain the corrected warning threshold;

[0040] Based on the revised warning threshold, real-time clinical data from the medical quality monitoring system is scanned to trigger medical risk warning signals.

[0041] On the other hand, the present invention provides an artificial intelligence-based medical quality monitoring system, comprising:

[0042] The target acquisition module is used to acquire the set of monitoring targets and the set of resource constraints of the medical quality monitoring system in real time.

[0043] The intensity analysis module is used to analyze the real-time cascading impact intensity of any two targets in the monitoring target set, and to analyze the chain fluctuation value of a single resource occupancy status change in the resource constraint set on the overall resource system.

[0044] The matrix generation module is used to identify the conflict correlation strength between different monitoring targets in the monitoring target set and generate a target conflict correlation matrix;

[0045] The target labeling module is used to mark target pairs with strong conflict coupling as a group when there are target pairs in the target conflict correlation matrix whose conflict correlation strength is greater than the correlation strength threshold.

[0046] The status identification module is used to identify a high-conflict-low-resource-pressure state when the real-time cascading impact intensity is greater than the cascading intensity threshold, the chain fluctuation value is greater than the fluctuation risk threshold, and there is a strongly conflicting coupled target group.

[0047] The early warning execution module is used to generate a dynamic adjustment coefficient for the early warning threshold based on preset conflict sensitivity factors and resource adequacy factors under high conflict and low resource pressure conditions, so as to correct the current early warning threshold and trigger medical risk early warning.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. By constructing a dual mechanism of collaborative optimization of conflict targets and dynamic resource adaptation, the inherent contradiction between sensitivity and false alarm rate in medical quality monitoring is effectively resolved. Unlike the existing static threshold early warning mode, a target conflict correlation matrix and a strong conflict coupled target group identification mechanism are introduced to accurately quantify the conflict intensity between monitored targets. Combined with real-time cascading influence intensity and chain fluctuation value analysis, a high conflict-low resource pressure state is determined under the resource constraint framework. When the system detects this state, the early warning threshold is dynamically adjusted based on the conflict sensitivity factor and resource margin factor, so that the system can automatically suppress the false alarm risk of secondary targets while ensuring the sensitivity of key targets. This solves the chronic problem of alarm fatigue and missed alarm risk caused by the fragmented optimization of targets in traditional systems.

[0050] 2. By monitoring the chain reaction between targets in real time through cascading impact intensity monitoring, and combining the chain fluctuation value to assess the stability of the resource system, the early warning decision is always anchored within the resource carrying capacity. When a high conflict-low resource pressure state is identified, the conflict sensitivity and resource adequacy are converted into threshold adjustment coefficients using mathematical mapping relationships. This enables the early warning sensitivity to increase with the intensity of conflict and converge with the increase of resource pressure. This avoids the false alarm storm when resources are overloaded and ensures the accurate capture of high-risk events when resources are sufficient, significantly improving the efficiency of medical resource utilization and the accuracy of risk prevention and control. Attached Figure Description

[0051] Figure 1 This is a flowchart of a medical quality monitoring method based on artificial intelligence according to the present invention;

[0052] Figure 2 This is a schematic diagram of the structure of a medical quality monitoring system based on artificial intelligence according to the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Example 1: Figure 1 This invention provides a medical quality monitoring method based on artificial intelligence, comprising:

[0055] S1. Real-time acquisition of the monitoring target set and resource constraint set of the medical quality monitoring system;

[0056] S2. Analyze the real-time cascading impact intensity of any two targets in the monitoring target set, and analyze the chain fluctuation value of a single resource occupancy status change in the resource constraint set on the overall resource system.

[0057] S3. Identify the conflict correlation strength between different monitoring targets in the monitoring target set and generate a target conflict correlation matrix;

[0058] S4. When there are target pairs in the target conflict correlation matrix whose conflict correlation strength is greater than the correlation strength threshold, they are marked as strongly conflicting coupled target groups.

[0059] S5. When the real-time cascading impact intensity is greater than the cascading intensity threshold, the chain fluctuation value is greater than the fluctuation risk threshold, and there is a strongly conflicting coupled target group, it is identified as a high conflict-low resource pressure state.

[0060] S6. Under the condition of high conflict and low resource pressure, generate a dynamic adjustment coefficient for the early warning threshold based on the preset conflict sensitivity factor and resource adequacy factor to correct the current early warning threshold and trigger a medical risk warning.

[0061] S1. Obtain the monitoring target set and resource constraint set of the medical quality monitoring system in real time. The specific implementation is as follows:

[0062] The real-time acquisition of the monitoring target set and resource constraint set of the medical quality monitoring system is implemented as follows: First, the medical quality monitoring system database is accessed. This database stores a list of monitoring target types predefined by the hospital's quality management committee. The list of monitoring target types exists in the form of a structured data table, containing two core entries: sensitivity optimization target entries and false alarm rate suppression target entries. Sensitivity optimization target entries record the identifiers of clinical risk event types that need to maximize the probability of correct warnings, such as postoperative infection risk events or acute kidney injury risk events. False alarm rate suppression target entries record the identifiers of clinical operation types that need to minimize the probability of false warnings, such as medication error warnings or fall risk warnings. The database extracts the complete list of monitoring target types using structured query language commands, ensuring that the target type definitions are consistent with the mandatory requirements of the hospital's clinical guidelines.

[0063] Based on the list of monitoring target types, real-time dynamic parameters are collected for each sensitivity optimization target item. For the sensitivity optimization target corresponding to the postoperative infection risk event identifier, its real-time parameter value is generated through the following process: The total number of currently hospitalized patients meeting the high-risk characteristics for postoperative infection is obtained from the hospital's electronic medical record system as the denominator; simultaneously, the number of correctly triggered postoperative infection warning events in the past 24 hours is extracted from the real-time warning log as the numerator; the percentage obtained by dividing the numerator by the denominator and multiplying by 100 is used as the real-time parameter value for this sensitivity optimization target. For the false alarm rate suppression target corresponding to the medication error warning identifier, its real-time parameter value is generated as follows: The total number of all medication operation records in the past 6 hours is obtained from the pharmacy management system as the denominator; simultaneously, the number of medication warnings erroneously triggered by the system in the same time period is extracted from the false alarm audit log as the numerator; the percentage obtained by dividing the numerator by the denominator and multiplying by 100 is used as the real-time parameter value for this false alarm rate suppression target. Finally, the real-time parameter values ​​of the above sensitivity optimization targets and false alarm rate suppression targets are stored according to target identifiers to form a monitoring target set data object.

[0064] The system synchronously reads the resource identifier list from the application programming interface (API) provided by the medical resource management platform. Maintained by the hospital resource scheduling center, the resource identifier list includes consumable resource identifier classification groups and occupied resource identifier classification groups. The consumable resource identifier classification groups enumerate codes for quantifiable medical supplies, such as specific drug inventory codes or batch codes for disposable medical consumables; the occupied resource identifier classification groups enumerate codes for shareable medical equipment or space with capacity limits, such as intensive care unit bed area codes or MRI equipment group codes. The platform transmits the resource identifier list in a lightweight data exchange format file via Hypertext Transfer Protocol (HTTP).

[0065] For consumable resource identifiers in the resource identifier list, a real-time data collection process is initiated. Taking vancomycin drug inventory codes as an example, the system queries the current pharmacy's actual remaining available inventory quantity through the drug inventory management system's application programming interface (API), and automatically scans the RFID tags on the drug shelves to obtain the physical count results. Taking disposable intravenous infusion needle batch codes as an example, the system obtains the current available warehouse inventory quantity through the material warehousing system's database view; this quantity value is updated based on the results of periodic inventory checks conducted by the inventory management software. For occupied resource identifiers in the resource identifier list, a status collection process is executed. Taking the coding of intensive care unit (ICU) bed areas as an example, the total number of open beds and occupied beds in the area are obtained through the real-time data interface of the hospital bed management system. The number of vacant beds is obtained by subtracting the number of occupied beds from the total number of open beds. This vacant bed number is then divided by the total number of open beds and multiplied by 100 to obtain a percentage value as the real-time available capacity. Similarly, taking the coding of magnetic resonance imaging (MRI) equipment groups as an example, the total number of available time slots and the number of occupied time slots are obtained through the medical equipment scheduling system. The number of available time slots is obtained by subtracting the number of occupied time slots from the total number of available time slots. This vacant time slot number is then divided by the total number of available time slots and multiplied by 100 to obtain a percentage value as the real-time available capacity. Finally, the real-time remaining available quantity data corresponding to consumable resource identifiers and the real-time available capacity data corresponding to occupied resource identifiers are merged and stored to form a resource constraint set data object.

[0066] Both the monitoring target set data objects and the resource constraint set data objects are encapsulated in a lightweight data exchange format and transmitted to subsequent analysis via an internal message queue. Real-time parameter values ​​for sensitivity optimization targets are retained to two decimal places, while real-time parameter values ​​for false alarm rate suppression targets are stored as percentage floating-point numbers. The real-time remaining available quantity of consumable resources is recorded as an integer unit, and the real-time idle available capacity of occupied resources is recorded as a percentage value. The data acquisition process is executed automatically every 5 minutes, and the last valid acquisition result is automatically activated from the local cache when the hospital's main information network is interrupted.

[0067] S2. Analyze the real-time cascading impact strength of any two targets in the monitoring target set, and analyze the cascading fluctuation value of a single resource occupancy status change in the resource constraint set on the overall resource system. Specifically, this is implemented as follows:

[0068] The operation to analyze the real-time cascading influence strength of any two targets in the monitoring target set is implemented as follows: Historical time-series records of the real-time parameter values ​​of the first and second targets are extracted from the data objects in the monitoring target set. These records contain parameter value sequences from the most recent six data acquisition cycles. Taking the postoperative infection risk sensitivity optimization target as the first target and the medication error false alarm rate suppression target as the second target as an example, the transmission rate of changes in the real-time parameter values ​​of the first target leading to changes in the real-time parameter values ​​of the second target is monitored. The transmission rate is calculated as follows: the absolute value of the difference between the real-time parameter values ​​of the first target within two adjacent acquisition cycles is obtained as the denominator; simultaneously, the absolute value of the difference between the real-time parameter values ​​of the second target within the same time interval is obtained as the numerator. The numerator is divided by the denominator to obtain the transmission rate value. For example, if the parameter value of the first target changes from 85.30 to 83.20 within 5 minutes, the absolute value of the difference is 2.10; if the parameter value of the second target changes from 12.50 to 14.80 during the same period, the absolute value of the difference is 2.30 ÷ 2.10 ≈ 1.095. The calculation rule for the impact amplitude is as follows: Obtain the absolute value of the difference between the real-time parameter values ​​of the second target within two adjacent acquisition cycles, and normalize this absolute value to eliminate the influence of differences in the dimensions of different targets. The transmission rate is calculated by monitoring the time difference between the start of the change in the first target parameter value and the start of the change in the second target parameter value; its value is the change in the second target parameter value per unit time. Finally, the absolute value of the transmission rate is multiplied by the normalized impact amplitude to obtain the real-time cascaded impact intensity. This intensity value only characterizes the severity of the impact, and its value is always positive.

[0069] It should be noted that the "cascade effect" analysis in this method is not applicable to all arbitrary pairs of objectives. Instead, it is based on prior medical knowledge, presupposing a set of objective pairs with potential clinical relevance or competing resource relationships. For example, "postoperative infection risk" and "antibiotic usage rate" are one set of related objective pairs, and "bed turnover rate" and "average length of stay" are another set. The system only calculates the strength of cascade effects for these presupposed, medically significant objective pairs, thereby ensuring the validity and clinical value of the calculation results.

[0070] The calculation of cascading fluctuation values ​​for target resource nodes within the resource constraint set is implemented as follows: Based on a predefined resource dependency mapping table in the hospital resource scheduling rule base, a set of associated resource nodes that have a direct dependency relationship with the target resource node is identified. Taking the intensive care unit bed area code as the target resource node as an example, its associated resource node set includes three items: nursing staff scheduling pool code, ventilator equipment group code, and emergency medicine inventory code. When the occupancy status of the target resource node changes, the decay rate of the real-time available capacity of each resource node in the associated resource node set is monitored. The absolute fluctuation range calculation rule is as follows: the real-time available capacity of the associated resource nodes before the target resource node's status change is recorded as the baseline value; after the status change stabilizes, the real-time available capacity is collected as the new value; and the absolute value of the difference between the new value and the baseline is calculated to obtain the absolute fluctuation amount of the resource node.

[0071] In this method, the chain fluctuation value is used to quantify the instability of the resource system, rather than simply judging resource scarcity. Dramatic resource fluctuations (regardless of whether they are increasing or decreasing) may indicate abnormalities in clinical workflows, such as sudden increases or decreases in patient numbers or emergency resource allocation; this instability itself is a potential signal of medical risk. Therefore, the absolute value is used in the calculation to capture the severity of the fluctuation. When the fluctuation value exceeds a preset fluctuation risk threshold, it indicates that the system is in an unstable state and requires attention.

[0072] The conduction rate calculation employs a fixed time window mechanism, with a fixed window width of two acquisition cycles (10 minutes). When the parameter value change is less than 0.10%, it is considered insignificant, and the conduction rate is directly recorded as zero. The resource dependency mapping table is updated periodically by the hospital's operations management department and stored as a corresponding structure of resource identifier codes and associated resource identifier code lists. The real-time idle available capacity conversion rule for consumable resources is: divide the real-time remaining available quantity by a preset total resource value and then multiply by 100, where the preset total resource value is determined based on material procurement records. All cascading impact intensity calculation results are stored in the form of a target-pair index matrix, with the row index representing the first target identifier and the column index representing the second target identifier; the chain fluctuation value calculation results are stored as a time series dataset categorized by resource identifier code. The trigger condition for changes in resource occupancy status is defined as: when the real-time idle available capacity of a target resource node changes by more than 5% within a single acquisition cycle, the chain fluctuation calculation is activated.

[0073] S3. Identify the conflict correlation strength between different monitoring targets in the monitoring target set, and generate a target conflict correlation matrix. The specific implementation is as follows:

[0074] It is worth noting that this step involves identifying the conflict correlation strength between pre-defined target pairs in the monitored target set that have a potential for conflict, particularly between sensitivity optimization targets and false alarm rate suppression targets, and generating a target conflict correlation matrix.

[0075] The operation of identifying the conflict correlation strength between any two targets in the monitoring target set and generating a target conflict correlation matrix is ​​implemented as follows: First, for each target pair consisting of a sensitivity optimization target and a false alarm rate suppression target in the monitoring target set, historical conflict alarm event records are extracted from the medical quality event database. The historical data time range is set to the most recent 30 days, with each day divided into 24 consecutive time windows. A sensitivity optimization target conflict alarm event is defined as an alarm record triggered when the real-time parameter value of the target is lower than a preset sensitivity threshold; a false alarm rate suppression target conflict alarm event is defined as an alarm record triggered when the real-time parameter value of the target is higher than a preset false alarm rate threshold. The co-occurrence frequency is calculated as follows: count the number of times a sensitivity optimization target conflict alarm event and a false alarm rate suppression target conflict alarm event occur simultaneously within the same time window, and divide this number by the total number of time windows to obtain the co-occurrence frequency value. For example, for a target pair consisting of a postoperative infection sensitivity optimization target and a medication error false alarm rate suppression target, alarms are triggered simultaneously 54 times in 720 time windows, and the co-occurrence frequency is 54 ÷ 720 = 0.075.

[0076] The relative difference between the real-time parameter values ​​of the sensitivity optimization target and the false alarm rate suppression target within the current time window is calculated synchronously. The current time window is defined as the most recent complete 60-minute interval, and the parameter snapshot values ​​recorded at the end of this time period are extracted from the real-time data stream of the medical quality monitoring system. The relative difference value is obtained through the following process: subtract the real-time parameter value of the false alarm rate suppression target from the real-time parameter value of the sensitivity optimization target to obtain the original difference value; divide the original difference value by the real-time parameter value of the sensitivity optimization target and multiply by 100 to convert it into a percentage form; finally, take the absolute value of this percentage value as the standardized difference value. For example, if the current parameter value of the postoperative infection sensitivity optimization target is 82.40, and the current parameter value of the medication error false alarm rate suppression target is 15.30, the original difference is 67.10, and the standardized difference value is |67.10÷82.40×100|=81.43%.

[0077] The conflict association strength is generated by weighting and summing the co-occurrence frequency and standardized difference values. A fixed coefficient model is used for weighting: the co-occurrence frequency weight coefficient A is set to 0.7, and the standardized difference value weight coefficient B is set to 0.3. The weighted sum calculation formula is: Conflict Association Strength = (Co-occurrence Frequency × Weight Coefficient A) + (Standardized Difference Value × Weight Coefficient B). The weight coefficient values ​​are determined by the hospital's quality control department based on historical conflict event analysis reports. Specifically, the weights are assigned by statistically analyzing the contribution of different factors to the actual conflict over the past six months, and stored in the system configuration library. Using the aforementioned example data: co-occurrence frequency 0.075 × 0.7 = 0.0525, standardized difference value 81.43 × 0.3 = 24.429, and the conflict association strength is 0.0525 + 24.429 = 24.4815.

[0078] The above calculation process is performed on all target pair combinations in the monitored target set. Target pair combinations include pairings between sensitivity-optimized targets, pairings between false alarm rate suppression targets, and cross-pairings between sensitivity-optimized targets and false alarm rate suppression targets. After generating a conflict association strength value for each target pair, a target conflict association matrix data structure is constructed. This matrix is ​​an N-order square matrix, where N equals the total number of targets in the monitored target set. Both row and column indices are arranged according to the storage order of the target identifiers in the set. The matrix element storage rule is as follows: the cell in the i-th row and j-th column stores the conflict association strength value of the target pair composed of the target at the row index and the target at the column index. For example, when the target at the row index is a postoperative infection target identifier and the target at the column index is a medication error target identifier, the corresponding cell stores 24.4815. The diagonal elements of the matrix are uniformly set to zero, indicating that the target has no conflict relationship with itself. The final generated complete target conflict association matrix is ​​stored in an in-memory database as a two-dimensional array structure.

[0079] Historical conflict alarm event records are acquired using a time window alignment mechanism, with each time window starting at the hour. Preset sensitivity thresholds are set according to clinical guidelines, such as a postoperative infection sensitivity threshold of 80.00; preset false alarm rate thresholds are set according to operational procedures, such as a medication error false alarm rate threshold of 10.00. When the medical quality event database has no historical records, the co-occurrence frequency is set to zero by default. When real-time parameter values ​​are missing, the most recently valid collected value is automatically used. All conflict correlation strength calculations retain four decimal places of precision. The weighting coefficient configuration file is reviewed quarterly by the quality control specialist, and a temporary revision process is initiated when the hospital's departmental structure is adjusted.

[0080] S4. When there are target pairs in the target conflict correlation matrix whose conflict correlation strength is greater than the correlation strength threshold, they are marked as strongly conflicting coupled target groups. The specific implementation is as follows:

[0081] When a target pair in the target conflict correlation matrix has a conflict correlation strength greater than the correlation strength threshold, the operation of marking it as a strongly conflict-coupled target group is implemented as follows: First, traverse all target pair units in the target conflict correlation matrix composed of sensitivity optimization targets and false alarm rate suppression targets. The traversal order is arranged in ascending order of the matrix row index. Each target pair unit contains the sensitivity optimization target identifier corresponding to the row index, the false alarm rate suppression target identifier corresponding to the column index, and the stored conflict correlation strength value. The initial correlation strength threshold is set to 20.00. This threshold is determined by analyzing the average conflict correlation strength value of actual medical conflict events that occurred in the hospital within the past three months. The specific calculation process is as follows: extract the conflict correlation strength values ​​of each target pair within 5 minutes before the event occurred in the historical conflict event records, calculate the arithmetic mean of these values ​​as the threshold benchmark value, and store it in the system configuration library. For example, if a target pair unit consisting of the postoperative infection sensitivity optimization target identifier and the medication error false alarm rate suppression target identifier is detected, its conflict correlation strength value is 24.4815.

[0082] If the conflict association strength value of a target pair cell is greater than the association strength threshold, an identifier addition operation is performed: the sensitivity optimization target identifier and the false alarm rate suppression target identifier in the target pair cell are added as paired elements to the conflict target identifier set. The conflict target identifier set is stored using a key-value pair data structure, where the key is the sensitivity optimization target identifier and the values ​​are a list of associated false alarm rate suppression target identifiers. For example, when the conflict association strength 24.4815 is greater than 20.00, the postoperative infection target identifier is stored as the key, and the medication error target identifier is stored as the first element of the value list. If the same sensitivity optimization target identifier appears multiple times, a new key is created only when it is added for the first time, and the associated false alarm rate suppression target identifiers are subsequently appended to the existing value list.

[0083] After completing the full traversal of the matrix, a merging operation is performed on the set of conflicting target identifiers: key-value pairs containing the same false alarm rate suppression target identifiers, or key-value pairs with different keys corresponding to the same sensitivity optimization target identifiers. The merging process is implemented based on graph structure analysis: each sensitivity optimization target identifier is considered as a left node, each false alarm rate suppression target identifier as a right node, and each key-value pair relationship is considered as an edge connecting the left and right nodes. Connected subgraphs are identified through a recursive adjacency access algorithm. Specifically, starting from an unvisited node, all nodes directly connected to it are visited, and then the directly connected nodes of these nodes are visited, until no new nodes are reachable. The set of target identifiers corresponding to all left and right nodes in each connected subgraph constitutes a strongly conflicting coupled target group. For example, there are three key-value pairs: key A associated with value list [X,Y], key B associated with value list [Y,Z], and key C associated with value list [W]. Then AXYBZ forms a connected subgraph, generating a strongly conflicting coupled target group {A,B,X,Y,Z}; CW forms an independent subgraph, generating another target group {C,W}.

[0084] The association strength threshold is updated quarterly by the hospital's information department, based on the most recent quarterly medical conflict event analysis report. When the conflict target identifier set is empty, an empty result set is returned, indicating no strongly conflicting coupled target group. If an isolated target identifier is found during the merging process (i.e., not connected to any other identifier), a minimal target group containing only a single sensitivity optimization target identifier and a single false alarm rate suppression target identifier is generated separately. The final generated strongly conflicting coupled target groups are stored in list form, with each list element being a sublist containing multiple target identifiers. The first part of each sublist contains the sensitivity optimization target identifier set, and the second part contains the false alarm rate suppression target identifier set. For example, the storage format of a strongly conflicting coupled target group is [[A,B,X,Y,Z],[C,W]], where A and B are sensitivity optimization target identifiers, and X,Y,Z,W are false alarm rate suppression target identifiers.

[0085] During traversal, if the conflict association strength value equals the association strength threshold, it is considered not to meet the condition and is skipped. The merge operation excludes duplicate target identifiers, meaning the same target identifier appears only once within the group. The termination condition for the recursive neighbor node access algorithm is set to stop when the current node has no unvisited directly connected nodes. All target identifier comparison operations use exact string matching rules, distinguishing between uppercase and lowercase characters.

[0086] S5. When the real-time cascading impact strength exceeds the cascading strength threshold, the chain fluctuation value exceeds the fluctuation risk threshold, and there is a strongly conflicting coupled target group, it is identified as a high conflict-low resource pressure state, specifically implemented as follows:

[0087] The operation to identify high-conflict, low-resource-pressure states is implemented as follows: The system first executes two independent comparison processes in parallel. The first comparison process acquires the set of real-time cascading impact intensity values ​​generated in step S2. This set contains the real-time cascading impact intensity values ​​corresponding to all target pairs in the monitored target set. Simultaneously, it reads the cascading intensity threshold from the system configuration library. The cascading intensity threshold is determined by calculating the 90th percentile value of the dataset of real-time cascading impact intensity values ​​five minutes before the occurrence of confirmed medical quality risk events within one calendar month (approximately 30 days) in the hospital's history. This threshold is dynamically updated monthly to ensure its timeliness. The comparison rule is: check whether each real-time cascading impact intensity value is strictly greater than the cascading intensity threshold; records that meet the condition are marked as true values, otherwise they are marked as false values.

[0088] The second comparison process synchronously acquires the cascading fluctuation value data object generated in step S2. This data object contains the cascading fluctuation values ​​corresponding to each resource node in the resource constraint set. Simultaneously, it reads the fluctuation safety threshold from the resource configuration file. The fluctuation risk threshold is determined by calculating the 90th percentile value of the cascading fluctuation value dataset five minutes prior to the occurrence of confirmed medical quality risk events within the past three months of the hospital's history. This threshold is used to identify the critical point at which the resource system enters an unstable state. The comparison rule is: check each cascading fluctuation value to see if it is strictly less than the fluctuation safety threshold; mark records that meet the condition as true values ​​and others as false values.

[0089] The system synchronously verifies the strongly conflicting coupled target group data structure generated in step S4. It extracts a target group sublist from the strongly conflicting coupled target group list and checks whether the first half of the sublist contains at least one sensitivity optimization target identifier and whether the second half contains at least one false alarm rate suppression target identifier. The verification rule is: traverse the first half of the sublist to confirm the existence of identifiers belonging to the sensitivity optimization target identifier set; traverse the second half of the sublist to confirm the existence of identifiers belonging to the false alarm rate suppression target identifier set. For example, in the strongly conflicting coupled target group [[A,B,X,Y]], where A and B are sensitivity optimization target identifiers and X and Y are false alarm rate suppression target identifiers, the verification result is true.

[0090] The system activates its status flag when three necessary conditions are met simultaneously: First, during the comparison process, there is at least one record with a real-time cascading effect strength value greater than the cascading strength threshold; second, during the comparison process, there is at least one record with a chain fluctuation value greater than the fluctuation safety threshold; and third, the strong conflict coupling target group verification result is true. The activation operation is as follows: set the high conflict-low resource pressure status flag bit to true in the system status register, and simultaneously record the trigger timestamp and the associated target group identifier. For example, when the real-time cascading effect strength value of the postoperative infection-medication error target pair is 9.20 greater than 8.50, the ventilator resource chain fluctuation value is 4.80% less than 5.00%, and the strong conflict coupling target group verification passes, the status flag is activated.

[0091] The cascading strength threshold and fluctuation risk threshold are updated monthly by the hospital's information department in collaboration with the clinical quality management department, based on a dataset generated from the analysis of medical quality event reports over the past 30 days. The fluctuation safety threshold is reviewed annually by the Medical Resource Management Committee and adjusted according to revisions to the hospital resource management system stability specifications, with temporary adjustments triggered when hospital infrastructure is upgraded. If the list of strongly conflicting coupled target groups is empty during verification, the system directly determines verification failure. Once activated, the status indicator remains valid until manually reset or the automatic reset conditions are met. Automatic reset conditions include: no real-time cascading impact strength value exceeding the cascading strength threshold is detected for three consecutive acquisition cycles, or the cascading fluctuation value returns to a safe range after the resource scheduling system completes optimization operations.

[0092] The parallel comparison process employs a dual-thread synchronous processing mechanism, with the first and second comparison threads starting and running independently simultaneously. The verification process begins after both comparison threads have completed, ensuring data timestamp consistency. All threshold comparison operations utilize a precise floating-point comparison algorithm with a precision control of 0.001. Upon activation of the status flag, an alarm notification process is triggered, sending a status code and timestamp information to the quality management center console via the hospital information system interface.

[0093] S6. Under high conflict and low resource pressure conditions, a dynamic adjustment coefficient for the early warning threshold is generated based on preset conflict sensitivity factors and resource adequacy factors to correct the current early warning threshold and trigger a medical risk warning. The specific implementation is as follows:

[0094] The operation of generating a dynamic adjustment coefficient for the early warning threshold under high conflict and low resource pressure conditions to correct the current early warning threshold and trigger medical risk early warnings is implemented as follows: When the high conflict and low resource pressure status flag in the system status register is true, the system immediately extracts the conflict sensitivity factor scalar and resource margin factor scalar associated with this status from the preset parameter configuration library. The conflict sensitivity factor scalar is determined by analyzing the proportion of delayed response cases in medical risk early warnings among high conflict events in the hospital's historical 6 months. The specific calculation process is as follows: Calculate the percentage of the total number of delayed response cases to the total number of high conflict events, divide this percentage by 100 to convert it to decimal form, and then multiply it by the adjustment coefficient 1.5 to obtain the conflict sensitivity factor scalar value. This value ranges from 0 to 0.3. For example, if the delayed response ratio of a certain department is 12.00%, then the conflict sensitivity factor scalar = 0.12 × 1.5 = 0.18. The resource margin factor scalar is determined by assessing the overall idle capacity percentage of the current resource constraint set. The specific calculation process is as follows: Calculate the weighted average of the real-time idle available capacity of all resource nodes in the resource constraint set. The weights are allocated according to the importance coefficient of the resource type. Divide this average by 100 to convert it to a decimal form, which is the resource margin factor scalar value. This value ranges from 0 to 1. For example, if the current overall idle capacity of the resource system is 63.50%, then the resource margin factor scalar = 0.635.

[0095] The system performs a natural logarithm operation on the conflict sensitivity factor scalar. This natural logarithm operation uses a Taylor series expansion algorithm implemented in a mathematical function library. The input parameter is the conflict sensitivity factor scalar value plus 1, and the calculation is performed 10 times, i.e., calculating ln(1 + conflict sensitivity factor scalar), with precision controlled to 0.0001. For example, when the conflict sensitivity factor scalar is 0.18, ln(1 + 0.18) = ln(1.18) ≈ 0.1655. Simultaneously, a square root operation is performed on the resource margin factor scalar, i.e., calculating the arithmetic square root of the resource margin factor scalar.

[0096] The dynamic adjustment coefficient for the warning threshold is obtained by multiplying the result of the natural logarithm calculation by the result of the square root calculation. The calculation formula is: the dynamic adjustment coefficient for the warning threshold equals the result of the natural logarithm calculation multiplied by the result of the square root calculation. Using the example data mentioned above: 0.1655 × 0.7968 ≈ 0.1319. This coefficient represents the proportional adjustment required for the warning threshold under high conflict and low resource pressure conditions, and the coefficient value ranges from 0 to 0.25.

[0097] The current warning threshold is read from the parameter storage area of ​​the medical quality monitoring system. This threshold is the baseline warning value used under normal system operation. The corrected warning threshold is obtained by multiplying this threshold by 1 and subtracting the difference between the current threshold and the dynamic adjustment coefficient. The correction formula is: Corrected warning threshold = Current warning threshold × (1 - Dynamic adjustment coefficient). For example, if the current postoperative infection warning threshold is 80.00 and the adjustment coefficient is 0.1319, then the corrected threshold = 80.00 × (1 - 0.1319) = 80.00 × 0.8681 = 69.448. This subtraction operation reflects the increased risk sensitivity requirements, making the warning threshold more stringent.

[0098] It should be noted that the calculation logic of the above-mentioned dynamic adjustment coefficient for the early warning threshold is based on the following design principles:

[0099] The role of the conflict sensitivity factor: An increase in the conflict sensitivity factor indicates that the system is in a high-conflict state, at which point the risk detection capability must be improved. Therefore, the component calculated by ln(1+conflict sensitivity factor) will increase, thereby guiding the system as a whole to adjust in a more sensitive direction.

[0100] The role of the resource adequacy factor: This method is specifically designed to utilize the resource adequacy factor. An increased resource adequacy factor indicates sufficient system resources, allowing for a potential increase in false alarms due to improved sensitivity. Therefore, in this case, the warning threshold should not be suppressed but rather allowed to be moderately relaxed (i.e., the threshold decreases while sensitivity increases). The calculated sqrt(resource adequacy factor) increases accordingly, reflecting this strategy of "supporting increased sensitivity when resources are abundant".

[0101] The overall effect of the coefficient: The final adjustment coefficient is the product of the conflict sensitivity component and the resource adequacy component. This means that the adjustment coefficient will only reach its maximum when the system is simultaneously in a "high conflict state" and "high resource adequacy" scenario, thereby maximally reducing the warning threshold and significantly improving monitoring sensitivity. When resources are scarce (small resource adequacy factor), even if the conflict is intense, the adjustment coefficient will be limited to a low level, thus avoiding excessive warnings and system overload caused by insufficient resources. This mechanism cleverly achieves the collaborative management goal of prioritizing system stability under high resource pressure and prioritizing the optimization of conflict targets under low resource pressure.

[0102] The clinical data stream of the medical quality monitoring system is scanned in real time based on a revised warning threshold. The scanning process involves comparing the real-time parameter value of each monitored target with a threshold value. When the real-time parameter value falls below the revised warning threshold, a medical risk warning signal is immediately triggered. The warning signal includes the target identifier, the real-time parameter value, the revised threshold, and the deviation value, and is transmitted to the monitoring terminal of the relevant clinical department through the hospital information system interface. For example, if the real-time parameter value of the postoperative infection target (68.30) is lower than the revised threshold (69.448), a warning signal is generated and the message "Postoperative infection risk: 68.30 < 69.448" is displayed.

[0103] The conflict sensitivity factor scalar is updated monthly, with the latest statistical data from the hospital's adverse event reporting system as the data source. The resource adequacy factor scalar is recalculated every minute to ensure it reflects the latest resource status. Early warning threshold corrections are completed within 5 seconds of status activation to ensure timely response. All calculations use double-precision floating-point format, and results are rounded to four decimal places. The weighting coefficients in the weighted average calculation of the resource adequacy factor are determined based on the importance of the resource type; for example, the weighting coefficient for intensive care unit beds is 0.5, for general ward beds it is 0.3, and for medical equipment it is 0.2.

[0104] Upon triggering a warning signal, a warning event log is generated, recording information such as the activation time, corrected parameter values, and triggering target. When the high conflict-low resource pressure status indicator is reset, the system automatically restores the warning threshold to its original value. The corrected threshold remains valid for a maximum of 2 hours, after which it automatically resets to prevent oversensitivity. Clinical departments must provide feedback on the processing results within 30 minutes of confirming the warning; the system then optimizes the calculation parameters of the conflict sensitivity factor scalar based on the feedback data.

[0105] The natural logarithm operation is implemented using the Taylor series expansion algorithm: The initial value of the accumulated sum is set to 0, and the iteration variable k is incremented from 1 to 10. The following operations are performed: Calculate the current term = (-1). (k+1) ×(Conflict Sensitivity Factor Scalar) k / k adds the current term to the sum; the final sum is an approximation of ln(1 + conflict sensitivity factor scalar). k represents the term number in the Taylor expansion, k=1 corresponds to the first term, k=2 corresponds to the second term, and so on up to the tenth term.

[0106] This embodiment achieves a leap from static thresholds to dynamic optimization in medical risk early warning by constructing a dual collaborative mechanism of conflict objectives and resource status. Existing technologies typically treat target conflict analysis and resource constraint management as two independent processes. However, this embodiment fuses real-time cascading impact intensity, chain fluctuation values, and target conflict correlation matrices into multi-dimensional data. Under high conflict and low resource pressure conditions, it generates early warning threshold adjustment coefficients through a nonlinear mapping between conflict sensitivity factors and resource adequacy factors.

[0107] The generation of the target conflict correlation matrix relies on dynamic analysis of the intensity of real-time cascading effects rather than historical static data. The calculation of the resource margin factor requires a comprehensive consideration of cascading fluctuation values ​​and real-time resource constraints. The generation of the threshold adjustment coefficient must simultaneously satisfy three conditions: conflict intensity exceeding the threshold, resource fluctuation safety, and the existence of a strongly conflict-coupled target group. A tight technical loop is formed between these steps, and their synergistic effect enables the system to automatically balance the conflict requirements of sensitivity and false alarm rate under resource-constrained conditions. This early warning mechanism based on multi-dimensional real-time data dynamic coupling demonstrates significantly better adaptability and reliability than traditional methods in actual clinical scenarios.

[0108] Example 2: Figure 2 A schematic diagram of an artificial intelligence-based medical quality monitoring system is provided. The artificial intelligence-based medical quality monitoring system includes:

[0109] The target acquisition module is used to acquire the set of monitoring targets and the set of resource constraints of the medical quality monitoring system in real time.

[0110] The intensity analysis module is used to analyze the real-time cascading impact intensity of any two targets in the monitoring target set, and to analyze the chain fluctuation value of a single resource occupancy status change in the resource constraint set on the overall resource system.

[0111] The matrix generation module is used to identify the conflict correlation strength between different monitoring targets in the monitoring target set and generate a target conflict correlation matrix;

[0112] The target labeling module is used to mark target pairs with strong conflict coupling as a group when there are target pairs in the target conflict correlation matrix whose conflict correlation strength is greater than the correlation strength threshold.

[0113] The status identification module is used to identify a high-conflict-low-resource-pressure state when the real-time cascading impact intensity is greater than the cascading intensity threshold, the chain fluctuation value is greater than the fluctuation risk threshold, and there is a strongly conflicting coupled target group.

[0114] The early warning execution module is used to generate a dynamic adjustment coefficient for the early warning threshold based on preset conflict sensitivity factors and resource adequacy factors under high conflict and low resource pressure conditions, so as to correct the current early warning threshold and trigger medical risk early warning.

[0115] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0116] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0117] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center containing one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0118] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0119] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0120] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0121] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0122] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0123] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0124] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A medical quality monitoring method based on artificial intelligence, characterized in that, include: S1. Real-time acquisition of the monitoring target set and resource constraint set of the medical quality monitoring system; S2. Analyze the real-time cascading impact intensity of any two targets in the monitoring target set, and analyze the chain fluctuation value of a single resource occupancy status change in the resource constraint set on the overall resource system. S3. Identify the conflict correlation strength between different monitoring targets in the monitoring target set and generate a target conflict correlation matrix; S4. When there are target pairs in the target conflict correlation matrix whose conflict correlation strength is greater than the correlation strength threshold, they are marked as strongly conflicting coupled target groups. S5. When the real-time cascading impact intensity is greater than the cascading intensity threshold, the chain fluctuation value is greater than the fluctuation risk threshold, and there is a strongly conflicting coupled target group, it is identified as a high conflict-low resource pressure state. S6. Under the condition of high conflict and low resource pressure, generate a dynamic adjustment coefficient for the early warning threshold based on the preset conflict sensitivity factor and resource adequacy factor to correct the current early warning threshold and trigger a medical risk warning.

2. The medical quality monitoring method based on artificial intelligence according to claim 1, characterized in that, Real-time acquisition of the monitoring target set and resource constraint set of the medical quality monitoring system, including: Extract a list of preset monitoring target types from the medical quality monitoring system database. The list of monitoring target types includes sensitivity optimization targets and false alarm rate suppression targets. The real-time parameter values ​​of the sensitivity optimization target and the false alarm rate suppression target are obtained to form a set of monitoring targets; The resource identifier list is retrieved from the medical resource management platform. The resource identifier list includes consumable resource identifiers and occupied resource identifiers. The real-time remaining available quantity corresponding to the resource consumption identifier and the real-time idle available capacity corresponding to the resource occupancy identifier are used to form a resource constraint set.

3. The medical quality monitoring method based on artificial intelligence according to claim 2, characterized in that, Analyze the real-time cascading impact strength of any two targets in the monitored target set, and analyze the cascading fluctuation values ​​of a single resource occupancy change in the resource constraint set on the overall resource system, including: For the first target and the second target in the monitoring target set, first determine whether there is a preset medical correlation between the two; if so, monitor the transmission rate and influence magnitude of the real-time parameter value change of the first target causing the real-time parameter value change of the second target, and generate the real-time cascaded influence intensity based on the transmission rate and influence magnitude. For a target resource node in the resource constraint set, identify a set of associated resource nodes that have a dependency relationship with the target resource node, and calculate the average absolute fluctuation of the real-time idle and available capacity of each resource node in the associated resource node set when the occupancy status of the target resource node changes, as the cascading fluctuation value.

4. The medical quality monitoring method based on artificial intelligence according to claim 3, characterized in that, The intensity of real-time cascaded effects is the product of the absolute value of the conduction rate and the magnitude of the effect, and is a non-negative scalar value.

5. A medical quality monitoring method based on artificial intelligence according to claim 3, characterized in that, Identify the conflict correlation strength between different monitoring targets in the monitoring target set, and generate a target conflict correlation matrix, including: For target pairs consisting of sensitivity optimization targets and false alarm rate suppression targets in the monitoring target set, extract the co-occurrence frequency of conflict alarm events between sensitivity optimization targets and false alarm rate suppression targets in historical data; Calculate the absolute value of the difference between the real-time parameter value of the sensitivity optimization target and the real-time parameter value of the false alarm rate suppression target within the current time window; The weighted sum of co-occurrence frequency and absolute value of difference is used as the conflict association strength; Traverse all target pair combinations in the monitored target set and generate a conflict association strength matrix with the target pair as the row and column index as the target conflict association matrix.

6. A medical quality monitoring method based on artificial intelligence according to claim 5, characterized in that, When there are target pairs in the target conflict correlation matrix whose conflict correlation strength is greater than the correlation strength threshold, they are marked as strongly conflict-coupled target groups, including: Traverse all target pair cells consisting of sensitivity optimization targets and false alarm rate suppression targets in the target conflict correlation matrix; If the conflict correlation strength of the target pair cell is greater than the correlation strength threshold, the sensitivity optimization target identifier and false alarm rate suppression target identifier in the target pair cell are added to the conflict target identifier set; Merge the target pair units that contain target identifiers with the same sensitivity optimization or the same false alarm rate suppression to generate a strongly conflict-coupled target group.

7. A medical quality monitoring method based on artificial intelligence according to claim 6, characterized in that, When the real-time cascading impact strength exceeds the cascading strength threshold, the chain fluctuation value exceeds the fluctuation risk threshold, and there is a strongly conflicting coupled target group, it is identified as a high-conflict-low-resource-pressure state, including: The system compares the real-time cascading effect intensity value with the cascading intensity threshold in parallel, and also compares the chain fluctuation value with the fluctuation safety threshold. Verify whether the strongly conflict-coupled target group contains sensitivity-optimized target identifiers and false alarm rate suppression target identifiers; When the real-time cascading impact strength value is greater than the cascading strength threshold, the chain fluctuation value is greater than the fluctuation safety threshold, and there are sensitivity optimization target identifiers and false alarm rate suppression target identifiers in the strongly conflict-coupled target group, the high conflict-low resource pressure state identifier is activated.

8. A medical quality monitoring method based on artificial intelligence according to claim 7, characterized in that, Under conditions of high conflict and low resource pressure, a dynamic adjustment coefficient for the early warning threshold is generated based on preset conflict sensitivity factors and resource adequacy factors to correct the current early warning threshold and trigger medical risk warnings, including: Obtain the scalar values ​​of the conflict sensitivity factor and resource margin factor corresponding to the activation of the high conflict-low resource pressure state identifier; Take the natural logarithm of the conflict sensitivity factor scalar, and simultaneously calculate the square root of the resource margin factor scalar; Multiply the natural logarithm result by the square root result to obtain the dynamic adjustment coefficient of the early warning threshold; Multiply the current warning threshold by the warning threshold dynamic adjustment coefficient to obtain the corrected warning threshold; Based on the revised warning threshold, real-time clinical data from the medical quality monitoring system is scanned to trigger medical risk warning signals.

9. A medical quality monitoring system based on artificial intelligence, used to implement the medical quality monitoring method based on artificial intelligence as described in any one of claims 1-8, characterized in that, include: The target acquisition module is used to acquire the set of monitoring targets and the set of resource constraints of the medical quality monitoring system in real time. The intensity analysis module is used to analyze the real-time cascading impact intensity of any two targets in the monitoring target set, and to analyze the chain fluctuation value of a single resource occupancy status change in the resource constraint set on the overall resource system. The matrix generation module is used to identify the conflict correlation strength between different monitoring targets in the monitoring target set and generate a target conflict correlation matrix; The target labeling module is used to mark target pairs with strong conflict coupling as a group when there are target pairs in the target conflict correlation matrix whose conflict correlation strength is greater than the correlation strength threshold. The status identification module is used to identify a high-conflict-low-resource-pressure state when the real-time cascading impact intensity is greater than the cascading intensity threshold, the chain fluctuation value is greater than the fluctuation risk threshold, and there is a strongly conflicting coupled target group. The early warning execution module is used to generate a dynamic adjustment coefficient for the early warning threshold based on preset conflict sensitivity factors and resource adequacy factors under high conflict and low resource pressure conditions, so as to correct the current early warning threshold and trigger medical risk early warning.

Images