An artificial intelligence-based nursing event analysis method and system

Abstract

The present application relates to the technical field of data mining, in particular to a nursing event analysis method and system based on artificial intelligence, comprising the following steps: collecting handover, signing, signature, alarm and document time and unifying time axis, marking responsibility site, merging case expression, screening consistent cases, mapping responsibility turning point and disposal track, and generating nursing event overall track analysis result. In the present application, the nursing event related records are rewritten along the unified time axis, the handover action, the execution action, the equipment alarm and the document record are connected into continuous responsibility site, and the verified case writing caliber is implemented semantic merging. The responsibility turning point and the bed abnormality are synchronously compared, and the analysis conclusion and the quality control review basis of the nursing event are closer to the on-site process.

Description

Technical Field

[0001] This invention relates to the field of data mining technology, and in particular to a nursing event analysis method and system based on artificial intelligence. Background Technology

[0002] Data mining technology refers to the technical field of data processing, correlation comparison, pattern summarization, and category identification around massive business records, process records, and result records. It typically includes determining data sources, extracting fields, cleaning records, setting labels, merging indicators, comparing conditions, frequency statistics, time series analysis, and summarizing rules. Overall, it often follows the method of collecting original records, unifying field definitions, removing duplicates and missing items, forming event labels based on preset standards, and then conducting corresponding analysis on the time sequence, personnel operations, handling process, and result descriptions of different records. The nursing event analysis method refers to the process of summarizing adverse nursing event reports, nursing records, doctor's orders, vital signs records, laboratory test results, shift handover records, and ward registration information for nursing events such as falls, medication errors, catheter dislodgement, pressure ulcers, and aspiration. This involves registering each event item according to its type, time of occurrence, location, patient details, procedure, treatment, and responsible parties. Then, based on the hospital's existing grading standards, reporting guidelines, and quality control criteria, the process involves manual classification, list comparison, selection of cause items, frequency statistics of similar events, and verification of the sequence of events to form an analysis of the nursing event content and its related matters.

[0003] Current nursing event analysis methods often summarize reports, records, and registration information by item, and then compare them item by item based on event type, time period, location, and responsibility. The operation process focuses on static field classification and manual selection. Handover confirmation, doctor's order receipt, execution signature, and equipment alarm are scattered. It is difficult to fall into a unified reference standard for the same event's responsibility transition and abnormal bedside rhythm. In actual operation, it is easy to have inaccurate corresponding treatment process, mixed retrieval of similar cases, and discontinuity of the responsibility chain. As a result, the analysis conclusions do not fit well with the on-site process, and the basis for quality control review is also easily scattered. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing an artificial intelligence-based nursing event analysis method.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a nursing event analysis method based on artificial intelligence, comprising the following steps: S1: Collect the handover confirmation time, doctor's order signing time, and execution signature time of nurses in the inpatient ward; collect the abnormal start time, abnormal recovery time, and alarm cancellation time of the bed monitoring equipment; collect the nursing document recording time; rewrite the landing point of each record according to a unified timeline; and establish a nursing event time-stamped chain. S2: Based on the nursing event time-stamped chain, arrange the order of handover check-out, handover confirmation, first round, doctor's order receipt, execution signature, abnormality initiation, abnormality recovery, and alarm event reporting, mark the responsibility gap, cross-shift jump point, and key treatment actions, and output the responsibility point sequence; S3: Based on the responsibility site sequence, read the ward verification case handover record and nursing document record, input the unified expression of the artificial intelligence semantic merging unit, rearrange the nursing case site order according to the same writing standard, and obtain nursing control benchmark data; S4: Based on the responsibility site sequence and the nursing control benchmark data, perform the first round of comparison for the responsibility gap landing point, the cross-shift jump landing point, and the abnormal start time relative to the first inspection position. Perform the second round of comparison for the abnormal recovery time, alarm cancellation time, and the relationship between key treatment actions. Use artificial intelligence to associate and merge consistent cases, simultaneously remove deviating cases, and generate an intelligent mapping list. S5: Based on the intelligent mapping list, read the bed location corresponding to the reported action, map the bedside responsibility turning point trajectory and treatment implementation trajectory of the current nursing event, and simultaneously write it into the nurse station terminal display page to generate the overall trajectory analysis result of the nursing event.

[0006] As a further aspect of the present invention, the nursing event time-stamped chain includes a time anchor identifier, an event time sequence label, and a source attribution label; the responsibility site sequence includes a responsibility acceptance segment, a treatment intervention segment, and an alarm connection segment; the nursing reference data includes semantic standard terms and case reference templates; the intelligent mapping list includes candidate case identifiers, trajectory matching labels, and deviation elimination marks; the overall trajectory analysis results of the nursing event in step five include a responsibility evolution profile, a treatment response profile, and a nursing event judgment conclusion.

[0007] As a further aspect of the present invention, the step of obtaining the nursing event time stamp chain specifically includes: S111: Collect the handover confirmation time, doctor's order signing time, and execution signature time of nurses in the inpatient ward using their handheld terminals, verify the time field format, establish a correspondence between the bed number and the doctor's order number, and generate a nursing operation time mapping table. S112: Obtain the abnormal start time, abnormal recovery time, alarm cancellation time and nursing record time of the bed monitoring equipment, compare the order of each time according to the nursing operation time mapping table, and establish an abnormal handling time sequence table; S113: Based on the abnormal handling time sequence table, associate it with the nursing operation time mapping table, rewrite the time points of handover confirmation, medical order receipt, execution signature, abnormal start, abnormal recovery, alarm cancellation, and document record to a unified time axis, and establish a nursing event time stamp chain according to bed number and event sequence.

[0008] As a further aspect of the present invention, the step of obtaining the responsible site sequence specifically includes: S211: Based on the nursing event time-stamped chain, extract the landing points of handover check-out, handover confirmation, first round, medical order receipt, execution signature, abnormal start, abnormal recovery, and alarm event reporting, check the order of the events along a unified time axis, and if the landing points are the same second, rearrange the order according to the rule that nursing operation precedes abnormal state and abnormal state precedes reporting record, and establish a time sequence point ranking list. S212: Obtain the preset shift time schedule, combine it with the time sequence location list, compare the interval between handover check-out and handover confirmation, the interval between handover confirmation and first round, and the interval between medical order receipt and execution signature, determine the responsibility gap landing point and cross-shift jump landing point based on the shift boundary time, mark the execution signature between the abnormal start and abnormal recovery as the key handling action, and generate a responsibility mark index. S213: Based on the responsibility marker index, associate the time sequence point list, write the responsibility gap landing point, cross-shift jump landing point, and key handling action into the corresponding event position, and connect the handover responsibility, shift takeover responsibility, patrol responsibility, execution responsibility, abnormal handling responsibility and reporting responsibility in chronological order to establish a responsibility point sequence.

[0009] As a further aspect of the present invention, the steps for obtaining the nursing control baseline data are specifically as follows: S311: Based on the responsibility site sequence, read the locally stored desensitized ward case handover records and nursing document records, check the correspondence by bed number, shift number, and event tag, split the handover statement and document statement, check the consistency of record time and responsibility position, check the consistency of statement source field, eliminate duplicate points, aggregate statements of the same bed and shift into continuous segments, and establish a case statement matching set; S312: Input the case statement arrangement set into the artificial intelligence semantic merging unit, retrieve synonyms, abbreviations, inverted statements and tense differences between handover statements and document statements, merge statements pointing to the same responsible event into unified event tags, rewrite the description time point into a unified time term, rewrite the disposal verb into a unified action word, rewrite the abnormal state into a unified state word, correct the subject reference relationship of the statement, and generate a writing caliber mapping table; S313: Based on the writing caliber mapping table, associate the case statement arrangement set with the responsibility site sequence, rewrite the nursing case site name according to the unified event label, rearrange the site order according to the order of handover responsibility, takeover responsibility, inspection responsibility, execution responsibility, abnormal handling responsibility and reporting responsibility, check the correspondence between the rearranged sites and responsibility positions, calculate the consistency index of site responsibility, map the handover statement sequence and the document statement sequence into a coaxial comparison structure, and establish nursing comparison benchmark data.

[0010] As a further aspect of the present invention, the formula for calculating the consistency index corresponding to the site responsibility is as follows: ; in, The consistency index corresponds to the responsibility of the representative site. The total number of liability events in representative cases, The indices representing the responsible events range from 1 to N. The statement mapping confidence for the nth responsible event (from the writing caliber mapping table). This represents the position of the nth responsible event after rearrangement. The standard responsibility position for the nth responsibility event is assigned as 1, 2, 3, 4, 5, 6 in the order of handover responsibility, takeover responsibility, inspection responsibility, execution responsibility, abnormal handling responsibility, and reporting responsibility. This represents the actual time interval between the nth and (n+1)th responsibility events. The standard time interval between the nth and (n+1)th incidents of responsibility (determined based on shift schedules or statistical averages of similar cases). This represents a preset minimum positive number used to avoid division by zero.

[0011] As a further aspect of the present invention, the step of obtaining the intelligent mapping list specifically includes: S411: Based on the responsibility site sequence and the nursing control baseline data, align the sites according to bed number, case number, and responsibility label, verify whether the responsibility gap falls within the baseline gap range, check whether the shift markers before and after the cross-shift jump point are continuous, compare whether the abnormal start time and the order of the first round of inspections are consistent, mark consistent items as coaxial items, mark offset items as difference items, and generate the first round of site marker clusters; S412: Based on the first round of site marker clusters, associate the nursing control baseline data, check whether the order of abnormal recovery time and alarm cancellation time corresponds, verify whether the key treatment action is located between the abnormality start and abnormal recovery, compare whether the relationship between the key treatment action and the alarm cancellation time is consistent, write the continuous relationship into the association label, write the reverse order relationship into the deviation label, and establish the second round of relationship screening clusters. S413: Based on the first round of site marker clusters and the second round of relationship filtering clusters, input the artificial intelligence association and merging unit to aggregate cases with the same bed, the same responsibility label, and the same sequence relationship. Remove cases with gap offset, shift break, reverse order restoration, and reverse order removal from the candidate set. Establish a mapping relationship between case number and responsibility label to generate an intelligent mapping list.

[0012] As a further aspect of the present invention, the steps for obtaining the overall trajectory analysis results of the nursing event are specifically as follows: S511: Based on the intelligent mapping list, read the bed location corresponding to the reported action, verify the correspondence between the bed number, case number, responsibility label and reporting time, merge consecutive reported actions of the same bed into the same group of nodes, rearrange the bed node positions according to time order, and establish a bed index. S512: Based on the bed index and the responsibility point sequence, map the handover responsibility, shift handover responsibility, patrol responsibility, execution responsibility, abnormal handling responsibility and reporting responsibility to the corresponding bed nodes, mark the bedside responsibility turning point according to the responsibility handover breakpoint, mark the handling implementation position according to the execution signature, abnormal recovery, alarm cancellation and reporting action position, and generate a dual-track bitmap by associating the two types of positions; S513: Based on the aforementioned dual-track bitmap, check the correspondence between the bedside responsibility turning point and the treatment implementation point and the bed, verify the time sequence of each node and the continuity of responsibility labels, synchronously write the bedside responsibility turning point trajectory and the treatment implementation trajectory into the nurse station terminal display page, associate the bed number, node position and trajectory segment identifier, and establish the overall trajectory analysis results of the nursing event.

[0013] An artificial intelligence-based nursing event analysis system includes: The time stamp acquisition module collects the handover confirmation time, medical order signing time, and execution signature time of nurses' handheld terminals in inpatient wards; it also collects the abnormal start time, abnormal recovery time, and alarm cancellation time of bed monitoring equipment; and it collects the nursing document recording time. It rewrites the landing point of each record according to a unified timeline and establishes a nursing event time stamp chain. The site indexing module, based on the nursing event time stamp chain, arranges the sequential positions of handover check-out, shift acceptance confirmation, first round of inspection, medical order receipt, execution signature, abnormality initiation, abnormality recovery, and alarm event reporting, marks the responsibility gap landing point, cross-shift jump landing point, and key treatment actions, and outputs the responsibility site sequence. The semantic compilation module reads the ward verification case handover records and nursing document records according to the responsibility site sequence, inputs them into the artificial intelligence semantic merging unit for unified expression, rearranges the nursing case site order according to the same writing standard, and obtains nursing control benchmark data. The clue screening module, based on the responsibility site sequence and the nursing control benchmark data, performs the first round of comparison for the responsibility gap landing point, the cross-shift jump landing point, and the abnormal start time relative to the first inspection location. It performs the second round of comparison for the abnormal recovery time, alarm cancellation time, and the relationship between key treatment actions. It uses artificial intelligence to associate and merge consistent cases, simultaneously eliminates deviating cases, and generates an intelligent mapping list. The analysis and aggregation module reads the bed location corresponding to the reported action based on the intelligent mapping list, maps the bedside responsibility turning point trajectory and treatment implementation trajectory of the current nursing event, and simultaneously writes it into the nurse station terminal display page to generate the overall trajectory analysis result of the nursing event.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, nursing event-related records are rewritten along a unified timeline, and handover actions, execution actions, equipment alarms, and document records are linked into continuous responsibility points. Then, the writing style of verified cases is semantically merged. By simultaneously comparing the dual clues of responsibility transition and bedside abnormalities, case screening no longer stops at the level of event name and item description, but shifts to the level of corresponding responsibility paths and handling trajectories. This narrows down irrelevant cases, fills in the gaps in the responsibility chain, and makes the nursing event analysis conclusions and quality control review basis closer to the actual situation. Attached Figure Description

[0015] Figure 1 This is a flowchart of the main steps of the present invention; Figure 2 This is a flowchart of the nursing event time-stamped chain acquisition process of the present invention; Figure 3 This is a flowchart of the process for obtaining the responsible site sequence in this invention; Figure 4 This is a flowchart illustrating the process of obtaining nursing control baseline data for this invention. Figure 5 This is a flowchart of the intelligent mapping list acquisition process of the present invention; Figure 6 This is a flowchart illustrating the process of obtaining the overall trajectory analysis results of nursing events according to the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0017] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0018] Please see Figure 1 An artificial intelligence-based nursing event analysis method includes the following steps: S1: Collect the handover confirmation time, doctor's order signing time, and execution signature time of nurses in the inpatient ward; collect the abnormal start time, abnormal recovery time, and alarm cancellation time of the bed monitoring equipment; collect the nursing document recording time; rewrite the landing point of each record according to a unified timeline; and establish a nursing event time-stamped chain. S2: Based on the nursing event time-stamped chain, arrange the order of handover check-out, handover confirmation, first round, doctor's order receipt, execution signature, abnormality initiation, abnormality recovery, and alarm event reporting, mark the responsibility gap, cross-shift jump point, and key treatment actions, and output the responsibility point sequence; S3: Based on the responsibility site sequence, read the ward verification case handover records and nursing document records, input the unified expression of the artificial intelligence semantic merging unit, rearrange the nursing case site order according to the same writing standard, and obtain nursing control benchmark data; S4: Based on the responsibility site sequence and nursing control benchmark data, the first round of comparison is performed on the responsibility gap landing point, the cross-shift jump landing point, and the abnormal start time relative to the first inspection location. The second round of comparison is performed on the abnormal recovery time, alarm cancellation time, and the relationship between key treatment actions. Artificial intelligence is used to associate and merge consistent cases, and deviating cases are simultaneously eliminated to generate an intelligent mapping list. S5: Based on the intelligent mapping list, read the bed location corresponding to the reported action, map the bedside responsibility turning point trajectory and treatment implementation trajectory of the current nursing event, and simultaneously write it into the nurse station terminal display page to generate the overall trajectory analysis results of the nursing event.

[0019] The nursing event time-stamped chain includes time anchor markers, event sequence labels, and source attribution labels; the responsibility location sequence includes responsibility acceptance segment, intervention segment, and alarm connection segment; the nursing baseline data includes semantic standard terms and case reference templates; the intelligent mapping list includes candidate case markers, trajectory matching labels, and deviation elimination markers; the overall trajectory analysis results of the nursing event in step five include responsibility evolution profile, intervention response profile, and nursing event judgment conclusion.

[0020] Please see Figure 2 Step S1 is as follows: S111: Collect the handover confirmation time, doctor's order signing time, and execution signature time of nurses in the inpatient ward using their handheld terminals, verify the time field format, establish a correspondence between the bed number and the doctor's order number, and generate a nursing operation time mapping table. The operation records of nurses' handheld terminals were extracted from the log table of the mobile nursing system's backend database in the inpatient ward. The data source for these operation records was mobile handheld terminals (PDAs) covered by the wireless network at the nurses' stations. Each PDA automatically recorded the operation type and timestamp when a nurse logged in. First, records with the operation type field value of "shift handover confirmation," "medical order receipt," and "medical order execution" were filtered from the data table. For each filtered record, the following fields were read: bed number, patient's unique identifier for this hospitalization, medical order number, operating nurse's employee number, and operation time. The original storage format of the operation time field was a Unix timestamp in milliseconds, such as "1712345678901." For subsequent comparison, this time field needed to be formatted and converted. Write an SQL query to convert Unix millisecond timestamps to a standard date and time format using database functions. The specific conversion rule is as follows: In SQL Server, use `DATEADD(ms, OperateTime%1000, DATEADD(s, OperateTime / 1000, '1970-01-01-08:00:00')` to obtain a time format like "2024-04-06-14:23:18.901" with milliseconds. After conversion, check that all time data is within a reasonable range; that is, the handover confirmation time should not be earlier than the start time of the current shift, and the order signing time should not be later than the order end time. Records with time values ​​earlier than January 1, 2024, or later than the current system time are marked as abnormal data and removed, and will not participate in subsequent processing. After completing the time format verification, establish the correspondence between data records using the bed number and order number as a composite key. The specific operation is as follows: The processed data is grouped according to bed number and order number. For the same order in the same bed, there may be multiple operation records; for example, one order may be signed or executed multiple times. Records within each group are sorted according to operation type and operation time, retaining the first record of each operation type in that group to ensure that each operation type corresponds to only one operation time point in each order. For example, for order "ORD-2024-04-05-008" in bed "12A-03", the handover confirmation time is "2024-04-06-08:05:12.000", the order signing time is "2024-04-06-08:10:35.000", and the execution signature time is "2024-04-06-09:20:41.000". This data is then organized into a structured table, where each row represents the three key operation time points corresponding to one order in one bed. The table contains the following fields: Bed Number, Order Number, Handover Confirmation Time, Order Receipt Time, and Execution Signature Time. The generated data table is named "Nursing Operation Time Mapping Table" and stored in the analysis database as the basis for subsequent steps.The mapping table contains nursing operation time data for 1,500 medical orders across 300 beds within 24 hours.

[0021] S112: Obtain the abnormal start time, abnormal recovery time, alarm cancellation time and nursing record time of the bed monitoring equipment, compare the order of each time according to the nursing operation time mapping table, and establish an abnormal handling time sequence table; Data is extracted from the central monitoring system workstation database and the nursing electronic documentation system. Monitoring equipment data originates from bedside monitors, each of which uploads patient vital signs data and alarm events to the central station in real time via a network interface. Alarm event records are read from the central station database. Each alarm record includes the bed number, abnormal start time, abnormal recovery time, and alarm clearing time. The abnormal start time is the point in time when the monitor detects vital signs parameters exceeding preset thresholds, such as heart rate > 120 beats / min or blood oxygen saturation < 90%. The abnormal recovery time is the point in time when the monitor detects parameters returning to the normal range. The alarm clearing time is the point in time when the nurse confirms the alarm by clicking the "clear alarm" button on the monitor or through the central station. The nursing document record time is read from the nursing electronic documentation system database; this time is the final save time when the nurse records the patient's condition, treatment measures, etc., in the electronic medical record. After extracting the above four types of time data, the time fields are uniformly converted to the same standard date and time format "YYYY-MM-DDHH:MM:SS.ms" as in step S111. For example, the monitoring system retrieves an alarm for decreased blood oxygen saturation in bed "12A-03": the abnormal start time is "2024-04-06-09:15:20.000", the abnormal recovery time is "2024-04-06-09:25:10.000", and the alarm clearing time is "2024-04-06-09:26:05.000". A record corresponding to this time period is retrieved from the nursing documentation, with a record time of "2024-04-06-09:30:00.000" and the record content "Patient's blood oxygen saturation decreased, improved after oxygen administration". Then, the retrieved monitoring alarm data is compared with the "Nursing Operation Time Mapping Table" generated in step S111. The comparison operation uses the bed number and the occurrence time window as the correlation basis. For each abnormal start time in the monitoring alarm data, the "Nursing Operation Time Mapping Table" is used to find records under the same bed number where the order receipt time and execution signature time fall within the time window of [abnormal start time - 30 minutes, abnormal recovery time + 30 minutes]. The purpose of the comparison is to determine the temporal relationship between nursing operations (order receipt, order execution) and changes in the patient's condition (abnormal start, recovery). For example, for the above alarm event, the "Nursing Operation Time Mapping Table" finds a record for bed "12A-03" with an order execution signature time of "2024-04-06-09:20:41.000", which falls between the abnormal start time "09:15:20" and the abnormal recovery time "09:25:10". The comparison results are then merged with all the obtained time points according to the bed number and the chronological order of the event to construct a new sequence. Each row of the sequence is centered on a complete abnormal event, recording all relevant time points that occurred before and after that event for that bed.The sequence includes the following fields: bed number, anomaly start time, anomaly recovery time, alarm cancellation time, nursing record time, medical order receipt time within the time period, and execution signature time. Finally, the sequence is grouped by bed number, and within each group, it is sorted according to the anomaly start time to generate an "anomaly handling time sequence table". As shown in Table 1, this table records the anomaly events and related time points for 5 beds on the morning of April 6.

[0022] Table 1. Time Series Table of Anomaly Handling (Example Excerpt): ; As shown in Table 1, the time series table of abnormal handling clearly shows the temporal correlation between nursing operations, document recording and monitoring alarms before and after each abnormal event, providing a foundation for the subsequent construction of a unified timeline.

[0023] S113: Based on the time sequence table of abnormal handling, and associated with the nursing operation time mapping table, the time points of handover confirmation, medical order receipt, execution signature, abnormal start, abnormal recovery, alarm cancellation, and document record are rewritten to a unified time axis, and a nursing event time-marked chain is established according to the bed number and the order of events. Based on the time series table of abnormal handling, and further linked to the nursing operation time mapping table, for each bed number, the time points of all types of events occurring within a specific time period are extracted and placed into a unified time coordinate system. Event types include: handover confirmation, doctor's order receipt, execution signature, abnormality initiation, abnormality recovery, alarm deactivation, and document record. Taking the event in bed "12A-03" on the morning of April 6, 2024 as an example, the handover confirmation time "08:05:12" and doctor's order receipt time "08:10:35" are obtained from the "Nursing Operation Time Mapping Table." The abnormality initiation time "09:15:20," execution signature time "09:20:41," abnormality recovery time "09:25:10," alarm deactivation time "09:26:05," and document record time "09:30:00" are obtained from the "Abnormal Handling Time Series Table." These seven time points are arranged in ascending order. The arrangement rule is to directly compare the numerical values ​​of the time points, with the smaller numerical value appearing first. For cases where the times are exactly the same, the original extraction order is temporarily maintained and will be processed in subsequent steps. The sorted time point sequence is as follows: T1=08:05:12 (handover confirmation), T2=08:10:35 (medical order receipt), T3=09:15:20 (abnormality start), T4=09:20:41 (execution signature), T5=09:25:10 (abnormality recovery), T6=09:26:05 (alarm deactivated), T7=09:30:00 (document record). Each time point in this sequence is treated as an "event endpoint," and each endpoint is assigned an attribute, including: occurrence time, event type, associated medical order number (if any), and associated abnormal event ID (if any). These event endpoints are then linked together in chronological order to form a "nursing event time-stamped chain" with time as the main axis. An independent time-stamped chain is generated for each bed. The data structure of the time-stamped chain can be represented as a list or array, where each element is a key-value pair containing a timestamp and event type. For example, the time-stamped chain for bed "12A-03" can be represented as: [{time: "08:05:12", event: "handover confirmation"}, {time: "08:10:35", event: "medical order receipt"}, {time: "09:15:20", event: "abnormal start"}, {time: "09:20:41", event: "execution signature"}, {time: "09:25:10", event: "abnormal recovery"}, {time: "09:26:05", event: "alarm deactivated"}, {time: "09:30:00", event: "document record"}]. The time-stamped chains for all beds in the hospital are summarized and stored in the database table "Nursing Event Time-Stamped Chain". The table fields include: bed number, event sequence number, event time, and event type.This time-stamped chain provides a precise time base and event sequence for all subsequent analyses.

[0024] Please see Figure 3 Step S2 is as follows: S211: Based on the nursing event time-stamped chain, extract the landing points of handover check-out, handover confirmation, first round, doctor's order receipt, execution signature, abnormal start, abnormal recovery, and alarm event reporting, check the order of the events along a unified time axis, and if the landing points are the same second, rearrange the order according to the rule that nursing operation precedes abnormal state and abnormal state precedes reporting record, and establish a time sequence point ranking list. From the nursing event time-stamped chain, specific event types are selected for each bed: handover check-out, handover confirmation, first round, doctor's order receipt, execution signature, abnormality initiation, abnormality recovery, and alarm event reporting. "Handover check-out" and "handover confirmation" correspond to the "handover confirmation" time in S111, but need to be distinguished based on the direction of the handover. During the data collection phase of S111, if the handover confirmation operation is initiated by the outgoing nurse, it is marked as "handover check-out"; if it is initiated by the incoming nurse, it is marked as "handover confirmation". The time point of "first round" needs to be extracted from nursing documentation records or other system logs, specifically referring to the time of the first physical examination or status assessment of the patient at that bed after the nurse takes over the shift. In this embodiment, it is assumed that this time has been obtained and stored in the time-stamped chain through other means, such as the time extracted from the "document record" type under the "nursing event time-stamped chain" based on the content keyword "first round". Taking bed "12A-03" as an example, the extracted event points and times are as follows: handover check-in "07:55:00", shift acceptance confirmation "08:05:12", first round of rounds "08:20:00", medical order receipt "08:10:35", execution signature "09:20:41", abnormality initiation "09:15:20", abnormality recovery "09:25:10", alarm event reporting "09:26:30". After extracting these events, their sequential positions are checked along a unified timeline. Sort these time points by value from smallest to largest to obtain the initial sequence: 07:55:00 (shift handover check-out), 08:05:12 (shift takeover confirmation), 08:10:35 (medical order receipt), 08:20:00 (initial rounds), 09:15:20 (abnormality initiation), 09:20:41 (execution signature), 09:25:10 (abnormality recovery), 09:26:30 (alarm event reporting). The check revealed that the medical order receipt time "08:10:35" is earlier than the initial rounds time "08:20:00," which is logically reasonable because nurses may handle urgent medical orders before conducting their initial rounds after taking over the shift. However, consider the scenario where two events occur within the same second, for example, the abnormality initiation and medical order execution signature times are both "09:20:41.500" and "09:20:41.800," but due to timestamp precision issues, they are recorded as the same second "09:20:41." At this point, a preset ranking rule needs to be applied: nursing operations (medical order execution signature) precede abnormal states (abnormal initiation). That is, if the execution signature and the abnormal initiation occur at the same second, the execution signature is considered to have occurred before the abnormal initiation. Another rule is: abnormal states precede reporting records. That is, if abnormal recovery and alarm reporting occur at the same second, abnormal recovery is considered to have occurred before reporting. In this example, no cases occurred at the same second, so the order remains unchanged. Based on the checked and rearranged order, a unique ranking number (1-based) is assigned to each event.For bed "12A-03", the sequence is as follows: 1. Shift handover check-out (07:55:00), 2. Shift handover confirmation (08:05:12), 3. Doctor's order receipt (08:10:35), 4. First round of rounds (08:20:00), 5. Abnormal event initiation (09:15:20), 6. Execution signature (09:20:41), 7. Abnormal event recovery (09:25:10), 8. Alarm event reporting (09:26:30). These events and their sequence information for each bed are compiled into a table to create a "Time-Sequence Sequence Ranking List". This table includes the fields: Bed Number, Event Sequence, Event Type, and Event Time.

[0025] S212: Obtain the preset shift time schedule, combine it with the time sequence location list, compare the interval between handover sign-out and handover confirmation, the interval between handover confirmation and first round, and the interval between medical order receipt and execution signature. Determine the responsibility gap landing point and cross-shift jump landing point based on the shift boundary time, mark the execution signature between the abnormal start and abnormal recovery as the key treatment action, and generate a responsibility mark index. Retrieve the pre-set shift schedule from the nursing scheduling system. This schedule defines the name, start time, and end time of each shift. For example, the day shift is from 08:00:00 to 16:00:00, and the night shift is from 00:00:00 to 08:00:00 and from 16:00:00 to 24:00:00. Taking bed "12A-03" as an example, its shift handover check-out time is "07:55:00," and its shift handover confirmation time is "08:05:12." Based on the shift schedule, the handover check-out time "07:55:00" falls within the night shift, and the shift handover confirmation time "08:05:12" falls within the day shift. Calculate the time interval between the handover check-out and the shift handover confirmation, i.e., from "07:55:00" to "08:05:12," which is 10 minutes and 12 seconds. Next, check the interval between the shift handover confirmation and the first round of rounds. The shift handover confirmation was at 08:05:12, and the first round of rounds was at 08:20:00, an interval of 14 minutes and 48 seconds. Next, the interval between the order receipt and execution signatures was checked. The order receipt time was 08:10:35, and the execution signature time was 09:20:41, an interval of 1 hour, 10 minutes, and 6 seconds. Responsibility was determined based on the shift boundary time. The shift boundary time was 08:00:00. Since the handover signature (07:55:00) was before the boundary, and the shift handover confirmation (08:05:12) was after the boundary, these two events belonged to different shifts, and the time interval of 10 minutes and 12 seconds crossed the shift boundary. Therefore, the time period between these two events, especially the boundary point "08:00:00," was marked as a "cross-shift jump point." This point represents the critical moment when nursing responsibility transfers from the night shift nurse (the handover nurse) to the day shift nurse (the incoming nurse). Next, any gaps in responsibility were identified. A responsibility gap is defined as a nursing procedure that should have been completed within a specified time but was not. For example, a hospital stipulates that the first round of rounds must be completed within 30 minutes of the shift start date. If the shift start date is 08:05:12 and the first round of rounds is 08:20:00, the interval is 14 minutes and 48 seconds, less than 30 minutes, therefore there is no responsibility gap for the first round of rounds. Similarly, if a doctor's order is signed for, it should be executed within 2 hours. If the order is signed for at 08:10:35 and the execution signature is signed at 09:20:41, the interval is 1 hour and 10 minutes, less than 2 hours, also therefore there is no execution gap. The determination of a responsibility gap requires a preset time threshold. If a time interval exceeds the preset threshold, that interval is marked as a "responsibility gap point." Then, key procedures are identified. The time sequence point ranking list is queried to find the abnormal start "09:15:20" and the abnormal recovery "09:25:10". All events between these two times are checked, and the execution signature "09:20:41" falls within this interval. Therefore, the execution signature event is marked as a "critical action", that is, a medical order executed during a period of abnormal patient condition.Finally, all the above marking information is integrated to generate a "responsibility marking index" for each bed. This index is a data structure that records the bed number, as well as all the marked special points on that bed (cross-shift transfer points, responsibility gap points, critical handling actions) and their associated events and specific times. For bed "12A-03", its responsibility marking index includes: a cross-shift transfer point (time: 08:00:00, associated events: shift handover check-out and shift takeover confirmation), and a critical handling action (event: execution signature, time: 09:20:41, associated abnormal event starting at 09:15:20).

[0026] S213: Based on the responsibility mark index, associate the time sequence position list, write the responsibility gap landing point, cross-shift jump landing point, and key handling action into the corresponding event position, and connect the handover responsibility, the takeover responsibility, the patrol responsibility, the execution responsibility, the abnormal handling responsibility and the reporting responsibility in chronological order to establish a responsibility position sequence; Based on the responsibility marker index and combined with the time-series position list, for bed "12A-03", the time-series position list already has 8 positions. The responsibility marker index marks a cross-shift transfer point and a critical treatment action. The cross-shift transfer point itself is not an independent event, but a time point (08:00:00) and a marker of responsibility transfer. This marker needs to be written between the two most relevant event positions. Specifically, a "cross-shift transfer" responsibility marker is inserted between position 1 (shift handover check-out, 07:55:00) and position 2 (shift handover confirmation, 08:05:12). This marker can be used as a special position, with its position set to 1.5 and its content being "cross-shift transfer responsibility". The critical treatment action (execution signature) is already at position 6; its attributes need to be strengthened, clarifying its responsibility type as "abnormal treatment responsibility". Simultaneously, it needs to be ensured that "execution responsibility" (i.e., ordinary medical order execution) and "abnormal treatment responsibility" overlap here. Next, each event in the site ranking list is assigned a responsibility type label based on its type. The specific rules are as follows: handover check-out corresponds to "handover responsibility," shift acceptance confirmation corresponds to "shift acceptance responsibility," the first round corresponds to "round responsibility," medical order receipt corresponds to part of "execution responsibility" (medical order receipt responsibility), execution signature during non-abnormal periods corresponds to "execution responsibility," during abnormal periods (marked as critical actions) corresponds to "abnormal handling responsibility," and the initiation, recovery, and reporting of abnormalities are combined to correspond to "reporting responsibility." For the initiation and recovery of abnormalities themselves, responsibility is not directly assigned; instead, the entire process from the initiation to the reporting of the abnormal event is considered a chain of responsibility. Events and responsibility labels are then reconnected in chronological order. For bed "12A-03", the sequence after concatenation is as follows: Position 1 (07:55:00) Shift handover check-out -> Shift handover responsibility; Position 1.5 (08:00:00) Cross-shift transfer landing point -> Shift responsibility transfer; Position 2 (08:05:12) Shift handover confirmation -> Shift handover responsibility; Position 3 (08:10:35) Medical order receipt -> Execution responsibility (receiving); Position 4 (08:20:00) Initial rounds -> Rounds responsibility; Position 5 (09:15:20) Abnormality initiation; Position 6 (09:20:41) Execution signature -> Abnormality handling responsibility (critical); Position 7 (09:25:10) Abnormality recovery; Position 8 (09:26:30) Alarm event reporting -> Reporting responsibility. These four points—abnormality initiation, critical handling, abnormality recovery, and reporting—are closely linked to form a complete "abnormality handling responsibility" unit. Ultimately, a sequence of responsibility points is generated, containing all beds, all events, and their corresponding responsibility labels. The table structure of this sequence includes: bed number, position, event time, event type, and responsibility label. This sequence clearly defines who should be responsible for what at each point in time.

[0027] Please see Figure 4 Step S3 is as follows: S311: Based on the responsibility site sequence, read the locally stored desensitized ward case handover records and nursing document records, check the correspondence by bed number, shift number, and event tag, split the handover statement and document statement, check the consistency of record time and responsibility position, check the consistency of statement source field, remove duplicate points, aggregate statements of the same bed and shift into continuous segments, and establish a case statement matching set; Verified case data from wards, which has undergone rigorous anonymization, is retrieved from a locally secure offline database. This data comprises two main parts: shift handover records and nursing documentation. Shift handover records are sourced from the nurse handover report system; each record includes the shift number, the incoming shift number, the handover nurse, the incoming nurse, the handover time, the incoming time, and the handover content in free text format. Nursing documentation records are sourced from nursing record sheets in the electronic medical record system; each record includes the recording time, the recording nurse, and the recording content (free text). Taking the case of bed "12A-03" on April 6, 2024, during the day shift (shift number D-2024-04-06) as an example: First, based on the bed's position in the responsibility location sequence, key time points on the responsibility timeline are determined. Then, handover statements related to the day shift are filtered from the handover records. For example, in the handover record between the night shift and day shift nurses, there is a passage: "Handover: Bed 12A-03, Patient Zhang San, vital signs were stable overnight, morning blood pressure 130 / 80, handover completed at 07:55. Close monitoring of blood oxygen levels is necessary." Selecting records from the nursing documentation for the day shift, such as: time "08:20:00", content "First round, patient alert, no discomfort reported."; time "09:30:00", content "Patient's blood oxygen saturation dropped to 88%, oxygen was administered at 5L / min, and retested at 95%." Next, perform the verification and splitting operations. Verify the correspondence between the statements and the responsibility site sequence. For example, if the first round time for position 4 in the responsibility site sequence is 08:20:00, the document statement "08:20:00" matches this time. Verify the consistency of the source field of the statements to ensure that the extracted statements are indeed from the nurse's records, and not from doctors or other personnel. For handover statements, they need to be split into finer units. The above handover text is split into: "Bed 12A-03", "Patient Zhang San", "Stable vital signs overnight", "Morning blood pressure 130 / 80", "Handover completed at 07:55", "Requires close monitoring of blood oxygen levels". Similarly, the document statements are also split into sentences. Then, remove duplicate entries. For example, if the handover record mentions "stable blood pressure" again, which is semantically redundant with a split statement, only that one statement is retained. Finally, perform statement aggregation. All the split sentence fragments from the same bed (12A-03) and the same shift (day shift D-2024-04-06) are aggregated into a continuous set of text fragments according to chronological order (if any) or logical order. This set constitutes the "case sentence arrangement set" for that bed and that shift. For the day shift patient in bed "12A-03", the statement arrangement set might include: [Handover statement 1: "Bed 12A-03"], [Handover statement 2: "Patient Zhang San"], [Handover statement 3: "Nighttime vital signs stable"], [Handover statement 4: "Morning blood pressure 130 / 80"], [Handover statement 5: "Handover completed at 07:55"], [Handover statement 6: "Blood oxygen saturation needs close monitoring"], [Document statement 1: "08:20:00 First round, patient alert, no discomfort reported"], [Document statement 2: "09:30:00 Patient's blood oxygen saturation dropped to 88%, after oxygen administration of 5L / min, it was retested at 95%"]. This arrangement set provides the raw material for subsequent semantic analysis.

[0028] S312: Input the case statement matching set into the artificial intelligence semantic merging unit, retrieve synonyms, abbreviations, inverted statements and tense differences between handover statements and document statements, merge statements pointing to the same responsible event into unified event labels, rewrite the description time point into unified time terms, rewrite the disposal verbs into unified action words, rewrite abnormal states into unified state words, correct the subject reference relationship of statements, and generate a writing caliber mapping table; Input the case statement array set into a pre-trained artificial intelligence semantic merging model. This model is based on the BERT architecture and has been fine-tuned on a large corpus of medical care texts. The model first tokenizes and tags the词性 of each input statement fragment. Then, it performs synonym and abbreviation term retrieval. For example, for "blood pressure 130 / 80" and "BP 130 / 80" in the statement, the model recognizes "blood pressure" and "BP" as synonyms and uniformly merges them into the standard term "blood pressure". For abbreviations, such as "respiratory failure" is unified as "respiratory failure". For inverted sentences, such as "After oxygen inhalation, the blood oxygen level rises to 95%, the patient" will be recognized by the model and adjusted to the normal word order "The patient's blood oxygen level rises to 95% after oxygen inhalation". For tense differences, such as "has handed over" and "will hand over" in "has" and "will", the model based on the context and the event occurrence time (07:55), uniformly rewrites the described time point into an absolute time or a relative standard term. For example, "has completed the handover at 07:55" in the merged "has" is bound to the time point and no longer serves as a separate status word. Next, event merging is performed. The model merges statements pointing to the same responsible event into a unified event label. For example, "Need to pay special attention to the blood oxygen situation" in the handover statement and "The patient's blood oxygen saturation drops to 88%" in the medical record statement, both of these statements point to the core event of "blood oxygen abnormality". The model calculates the semantic similarity and merges both of them under the event label "blood oxygen abnormality". Then, the verbs and status words are rewritten. The model uniformly rewrites verbs describing treatment actions, such as "give", "inhale", "apply", etc., into "perform". For example, "Give oxygen at 5L / min" in the "give" is rewritten as "perform". Words describing abnormal states, such as "drop", "low", "unstable", etc., are uniformly rewritten as "abnormal". For example, "drop" in "The blood oxygen saturation drops" is rewritten as "abnormal". At the same time, the reference relationship of the sentence subject is corrected. The model contains a reference resolution module that can recognize that "patient", "this patient", "Zhang San" all point to the same entity, and when generating the mapping, uniformly uses the standard term "patient". After all the above processes, the model generates a standardized output for each original statement fragment. These standardized outputs form a one-to-one mapping with the original statement fragments. Organize all the mapping relationships into a table, which is the "writing caliber mapping table". Table 2 shows some examples of mapping relationships.

[0029] Table 2 Writing Caliber Mapping Table: ; As shown in Table 2, the mapping table unifies the colloquial and diverse original records into a standardized expression with clear structure and consistent terms, providing a bridge for subsequent construction of structured data.

[0030] S313: Based on the written caliber mapping table, associate the case statement allocation set with the responsibility position sequence, rewrite the nursing case position names according to the unified event label, rearrange the position order according to the sequence of handover responsibility, shift handover responsibility, rounds responsibility, execution responsibility, abnormal handling responsibility, and reporting responsibility, verify the correspondence between the rearranged positions and responsibility positions, and use the formula: ; Consistency indices corresponding to site responsibility were calculated, and the handover statement sequence and the document statement sequence were mapped to a coaxial control structure to establish nursing control benchmark data; among which... The consistency index corresponds to the responsibility of the representative site. The total number of liability events in representative cases, The indices representing the responsible events range from 1 to N. The statement mapping confidence for the nth responsible event (from the writing caliber mapping table). This represents the position of the nth responsible event after rearrangement. The standard responsibility position for the nth responsibility event is assigned as 1, 2, 3, 4, 5, 6 in the order of handover responsibility, takeover responsibility, inspection responsibility, execution responsibility, abnormal handling responsibility, and reporting responsibility. This represents the actual time interval between the nth and (n+1)th responsibility events. The standard time interval between the nth and (n+1)th incidents of responsibility (determined based on shift schedules or statistical averages of similar cases). This represents a preset minimum positive number to avoid division by zero; First, based on the writing guidelines mapping table, and combined with the case statement allocation set and responsibility site sequence, the site names in the nursing cases are rewritten. For the case of bed "12A-03", the event points in its responsibility site sequence, such as "shift handover confirmation" at position 2, have their corresponding standardized name "shift handover event" found in the "writing guidelines mapping table". Therefore, the site name is rewritten as "shift handover event". Similarly, "first round" at position 4 is rewritten as "round event", and "execution signature" at position 6 is rewritten as "execution event", and they are associated with the responsibility for handling abnormalities. After completing the name rewriting, the site order needs to be rearranged according to the predefined responsibility order. The predefined responsibility order is: handover responsibility, shift handover responsibility, round responsibility, execution responsibility, abnormality handling responsibility, and reporting responsibility. For the case of bed "12A-03", the original order of events is: handover event (position 1), takeover event (position 2), execution event - reception (position 3), inspection event (position 4), anomaly initiation (position 5), execution event - anomaly handling (position 6), anomaly recovery (position 7), and reporting event (position 8). During rearrangement, events of the same responsibility type are grouped together. However, the events themselves are fixed in time; the purpose of rearrangement is for comparison, not to change the actual occurrence time. Here, "rearrangement" refers to establishing a logical order index for subsequent comparison, not physically changing the sequence. In actual execution, the original positions are aligned with the standard responsibility order in a new mapping structure. For example, in the standard sequence, the first position is the handover responsibility, corresponding to the original position 1; the second position is the takeover responsibility, corresponding to the original position 2; the third position is the rounds responsibility, corresponding to the original position 4; the fourth position is the execution responsibility, corresponding to the original positions 3 (receiving medical orders) and 6 (abnormal handling, but here the execution responsibility is considered together, as abnormal handling is a special form of execution); the fifth position is the abnormal handling responsibility, corresponding to the original position 6; and the sixth position is the reporting responsibility, corresponding to the original position 8. When verifying the correspondence, it is necessary to confirm that each rearranged position can find a unique and correct correspondence in the original sequence.

[0031] Then, the formula is applied to calculate the consistency index corresponding to site responsibility. The parameters in the formula need to be obtained from the data. For the case of bed "12A-03", This represents the total number of liability events in this case. According to the liability location sequence of S213, the main liability events involved in this case fall into six categories: shift handover responsibility, shift takeover responsibility, patrol responsibility, execution responsibility, abnormal handling responsibility, and reporting responsibility. Therefore... . These events are indexed from 1 to 6. It is the first The statement mapping confidence score for each responsibility event is obtained from the writing style mapping table. For the "shift handover event," its mapping confidence score is... The value is 0.95 (output by the model). For the "succession event", The value is 0.98. (Regarding the "inspection incident") The value is 0.96. (Execution Event - Medical Order Receipt Section) The value is 0.97. "Responsibility for Handling Abnormal Situations" (Executive Signature) The value is 0.99. "Reporting responsibility" It is 0.98. It is the first The reordering of responsibility events. This reordering is a logical order based on the standard responsibility sequence, i.e., handover responsibility. Succession responsibility Inspection responsibility , execution responsibility Responsibility for handling abnormal situations Report responsibility . It is the first The standard order of responsibility for each incident is predetermined and consistent with... The definition is consistent, that is , , , , , Therefore, for all , . It is the first The first incident of responsibility and the first The actual time interval between each responsible event. According to the time-stamped chain of S113, (Interval between shift handover and handover) = 08:05:12 - 07:55:00 = 612 seconds. (Interval between shift handover and patrol) = 08:20:00 - 08:05:12 = 888 seconds. (The interval between rounds and execution refers to the time point of the execution responsibility event, i.e., the interval between receiving the medical order at 08:10:35 and the rounds at 08:20:00. Note: The order of responsibility and the order of time may not be completely consistent. The time interval calculation here should be based on the actual time sequence, not the order of responsibility. The formula...) It should be defined as the interval between responsibility events arranged in chronological order. Therefore, the time points corresponding to these 6 responsibility events need to be rearranged in chronological order first. The actual time sequence is: shift handover (07:55:00), shift takeover (08:05:12), medical order receipt (08:10:35), rounds (08:20:00), abnormal handling (09:20:41), and reporting (09:26:30). Therefore, (Shift handover - Shift takeover) = 612 seconds (Shift handover - Doctor's orders received) = 323 seconds (Medical order receipt - rounds) = 565 seconds (Inspection - Anomaly Handling) = 3621 seconds (Abnormal handling - reporting) = 349 seconds. This is a standard time interval, obtained from a preset shift schedule or by averaging similar cases. In this embodiment, The standard shift handover interval is set at 600 seconds (10 minutes). The standard interval for processing the first medical order after taking over the shift is set at 300 seconds (5 minutes). The standard interval for the first round of rounds after receiving the doctor's orders is set at 1200 seconds (20 minutes). The standard interval between the start of an inspection and the occurrence of an anomaly is not defined, but for calculation purposes, we take the statistical average of similar cases, assuming it to be 3600 seconds. The standard interval between handling an anomaly and reporting it is set to 300 seconds. This is a preset minimum positive number, taken as 0.01, to avoid division by zero. Substitute the above value into the formula for calculation.

[0032] First, calculate the first part: numerator = Denominator = Therefore, the ratio of the first half is 1. Next, calculate the square root of the value in the second half. Calculate the square of each term: First item ,square ; Second item ,square ; Third item ,square ; Fourth item ,square ; Fifth item ,square ; The sum of these 5 items = , divided by ,get Taking the square root yields approximately .final, .

[0033] The results indicate that although the positional correspondences are completely consistent (the first half is 1), the consistency index suffers from a significant deviation between the actual time interval and the standard time interval (especially the third item, the abnormal interval between inspection and execution). The value is only 0.165. The lower this value, the greater the deviation between the actual process and the standard process. This calculation process maps the originally separate handover statement sequences and document statement sequences onto the same axis for comparison using a unified quantitative indicator. Finally, the calculated... The values ​​and their rearranged loci correspondences are stored as part of the nursing control baseline data, providing a basis for subsequent anomaly detection and quality assessment. The advantage of the formula lies in the introduction of confidence weights. The time deviation penalty item can comprehensively assess the degree to which nursing records conform to standard procedures in both semantic and temporal dimensions.

[0034] Please see Figure 5 Step S4 is as follows: S411: Based on the responsibility site sequence and nursing control baseline data, align the sites according to bed number, case number, and responsibility label, verify whether the responsibility gap falls within the baseline gap range, check whether the shift markers before and after the cross-shift jump point are continuous, compare whether the abnormal start time and the order of the first round of rounds are consistent, mark consistent items as coaxial items, mark offset items as difference items, and generate the first round of site marker clusters; Retrieve the responsibility site sequence and nursing baseline data, taking the case of bed "12A-03" on the day shift on April 6, 2024, as an example (case number C-2024-04-06-12A03). First, align the information from the two data sources according to bed number "12A-03", case number "C-2024-04-06-12A03", and responsibility labels (such as shift handover responsibility, shift takeover responsibility, etc.). The responsibility site sequence records each responsibility event in this case and its occurrence time. The nursing baseline data includes standardized event names and standard time intervals obtained through semantic analysis. And the relevant parameters obtained through formula calculation.

[0035] Next, the verification process will be carried out.

[0036] The first verification step: Verify whether the responsibility gap falls within the baseline gap range. The "responsibility gap" marked in step S212 is determined based on a time threshold. For example, if a hospital stipulates that the first round of inspections after taking over a shift must not exceed 30 minutes, this "30 minutes" is the upper limit of the baseline gap range. In this case, the shift take-over is confirmed at 08:05:12, and the first inspection is at 08:20:00, an interval of 14 minutes and 48 seconds, which is less than 30 minutes. Therefore, the gap is not marked in S212. In another case, if the shift take-over is confirmed at 08:05 and the first inspection is at 08:45, an interval of 40 minutes, then S212 would mark this point as the "inspection responsibility gap." Now, verify this gap. The baseline gap range is [0, 1800] seconds (i.e., within 30 minutes). The actual interval of 40 minutes (2400 seconds) exceeds the upper limit, so this marking is accurate. This gap is determined to be a "consistent item."

[0037] The second verification step: Checking whether the shift markers before and after the cross-shift jump point are continuous. For bed "12A-03", S212 marks a cross-shift jump point at 08:00:00. Its preceding event, shift handover check-out (07:55:00), belongs to the night shift, and the subsequent event, shift handover confirmation (08:05:12), belongs to the day shift. From the shift markers, it can be seen that the jump is directly from the night shift to the day shift, without interference from other shifts (such as the beginning and end shifts, or on-call shifts), so the shift markers are continuous. This item is judged as a "consistent item". If the preceding event is a night shift, and the event after the jump is still a night shift (for example, the shift handover confirmation also occurs during the night shift), then the markers are not continuous and will be judged as an "offset item".

[0038] The third verification step: Compare the time of the abnormality's onset with the order of the initial rounds. In this case, the abnormality's onset time was 09:15:20, and the initial rounds were at 08:20:00. The abnormality began after the initial rounds. Nursing routines require nurses to first round the patient upon taking over the shift to establish a baseline understanding of the patient's condition before promptly identifying subsequent abnormalities. Therefore, the abnormality's onset occurring after the initial rounds is logically consistent, and this item is classified as a "consistent item." Conversely, if the abnormality's onset time is earlier than the initial rounds, it indicates that the patient's condition had already changed before the nurse first saw the patient. This constitutes a logical error in the nursing process and will be marked as a "deviation item."

[0039] After all verifications are completed, all verification results are summarized. Sites or relationships judged as "consistent" are labeled "coaxial." Sites judged as "offsets" are labeled "differences." This set of labeled sites constitutes the "first-round site label cluster" for this case. For example, for case C-2024-04-06-12A03, its first-round site label cluster includes: no-responsibility gap landing points (considered consistent), cross-shift jump landing points (coaxial), and the relationship between the start of anomalies and the order of patrols (coaxial).

[0040] S412: Based on the first round of site marker clusters, associate the nursing control baseline data, check whether the order of abnormal recovery time and alarm cancellation time corresponds, verify whether the key treatment action is located between the abnormality onset and abnormal recovery, compare whether the relationship between the key treatment action and the alarm cancellation time is consistent, write the continuous relationship into the association label, write the reverse order relationship into the deviation label, and establish the second round of relationship screening clusters. Based on the initial locus marker clusters, the nursing control baseline data were re-linked to conduct a deeper round of verification of the relationships between events in the cases. Case C-2024-04-06-12A03 in bed "12A-03" was still used as an example for analysis.

[0041] The first check is to verify the correspondence between the abnormal recovery time and the alarm cancellation time. From the time-stamped chain in this case, we know the abnormal recovery time is "09:25:10" and the alarm cancellation time is "09:26:05". In clinical practice, nurses will only go to the bedside or cancel the alarm on the system after the patient's vital signs return to normal (abnormal recovery). Therefore, the recovery time should precede the cancellation time. Checking their order, the recovery time is indeed earlier than the cancellation time; the correspondence is correct. This relationship is marked as "continuous relationship".

[0042] The second verification step: Whether the critical intervention occurred between the onset and recovery of the abnormality. Step S212 has marked the execution signature "09:20:41" in this case as a critical intervention. The onset of the abnormality is "09:15:20," and the recovery is "09:25:10." The critical intervention time "09:20:41" does indeed fall between these two time points, meeting the definition of "intervention performed during the patient's abnormality." This verification passes, and this relationship is also marked as a "continuous relationship."

[0043] The third comparison: whether the critical intervention and alarm cancellation times are consistent. The critical intervention was at "09:20:41," and the alarm cancellation was at "09:26:05." The intervention came first, followed by the cancellation, which aligns with the logic of intervention first, and then deactivation after the patient's condition stabilizes. If the alarm cancellation time is earlier than the critical intervention time, it indicates that the nurse cancelled the alarm before performing the critical intervention, which is an abnormal procedure. In this case, the relationship is consistent and is marked as "continuous relationship." If any of the above checks results in a reversed order, such as the abnormal recovery time being later than the alarm cancellation time, this relationship will be written into the "deviation label." After completing all the above checks, all event relationships marked as "continuous relationship" are aggregated to form a "secondary round relationship screening cluster." For case C-2024-04-06-12A03, its "secondary round relationship screening cluster" includes three sets of relationships: (abnormal recovery - alarm cancellation) continuous, (critical intervention - abnormal interval) continuous, and (critical intervention - alarm cancellation) continuous.

[0044] S413: Based on the first round of site marker clusters and the second round of relationship screening clusters, input the artificial intelligence association and merging unit to aggregate cases with the same bed, the same responsibility label, and the same sequence relationship. Remove cases with gap offset, shift break, reverse order restoration, and reverse order removal from the candidate set. Establish a mapping relationship between case number and responsibility label to generate an intelligent mapping list. The first-round site marker clusters and the second-round relationship filtering clusters are fed as joint input into the AI-based association and merging unit. The core of this unit is a graph-based merging algorithm. First, all cases are grouped according to bed number. Within the same bed group, they are further grouped according to responsibility labels (such as "inspection responsibility" and "abnormal handling responsibility"). Then, within groups with the same bed and the same responsibility label, they are further aggregated according to the temporal relationship characteristics presented in the "first-round" and "second-round" marker clusters. Temporal relationship characteristics include combinations of labels such as "coaxial," "difference," "continuity," and "deviation." For example, all cases marked as "coaxial" in "abnormality initiation and inspection sequence" and "continuous" in "critical handling and abnormal interval relationship" can be aggregated into one category, representing cases that conform to the process specifications. The association and merging unit performs a filtering operation, marking all cases with "offset items" or "deviation labels." Specifically, these include cases with shift gaps (i.e., the gaps marked as "differences" in S411), cases with shift breaks (discontinuous handover records), cases with reversed recovery order (abnormal recovery time later than alarm clearance time), and cases with reversed clearance order (alarm clearance earlier than critical treatment). These cases are considered to have flaws in nursing procedures or responsibility fulfillment and are removed from the candidate set of "excellent cases." Taking an analysis of 20 cases in the same bed as an example, after the first and second rounds of screening, it was found that 3 cases had shift gaps (initial rounds exceeding the time limit), 2 cases had shift breaks (discontinuous handover records), and 1 case had reversed clearance order (alarm clearance time earlier than the execution time of the nursing record). These 6 cases were removed from the candidate set. The remaining 14 cases performed consistently in all verifications, forming the core candidate set. Finally, for each case in the core candidate set, a precise one-to-one mapping relationship was established according to its case number and responsibility label. For example, for case C-2024-04-06-12A03, the mapping relationship is: {Case Number: C-2024-04-06-12A03, Responsibility Tag Set: [Shift Handover Responsibility, Shift Intake Responsibility, Inspection Responsibility, Execution Responsibility, Abnormal Handling Responsibility, Reporting Responsibility], Temporal Relationship Characteristics: [Coaxial, Continuous, ...]}. These mapping relationships for all selected cases are summarized to generate the final "Intelligent Mapping List." This list provides high-quality, standardized sample data for subsequent nursing quality analysis and trajectory display.

[0045] Please see Figure 6 The S5 steps are as follows: S511: Based on the intelligent mapping list, read the bed location corresponding to the reported action, verify the correspondence between the bed number, case number, responsibility label and reporting time, merge consecutive reported actions of the same bed into the same group of nodes, rearrange the bed node positions according to time sequence, and establish a bed index. Starting with the intelligent mapping list, which contains information on all high-quality cases that have passed the screening, records related to "reporting responsibility" are extracted for each case in the list. Reporting responsibility corresponds to the time point when the alarm event was reported. First, the bed number and corresponding reporting time point for each case in the list are read. For example, the following information is obtained from the list: Bed "12A-03", Case C-2024-04-06-12A03, Reporting time "09:26:30"; Bed "12A-03", Case C-2024-04-07-12A03 (a case on another date), Reporting time "15:10:00". Next, the correspondence between the bed number, case number, responsibility tag, and reporting time is verified to ensure that each reporting record can be uniquely associated with a specific case and a specific bed. For example, confirm that the time point "09:26:30" does indeed originate from the "reporting responsibility" event in case C-2024-04-06-12A03. After verification, process all reporting actions for each bed. For the same bed, there may be multiple cases, each with one or more reports. Sort all reporting time points for the same bed in chronological order. For example, for bed "12A-03," there are two reporting points: 2024-04-06-09:26:30 and 2024-04-07-15:10:00. Then, group these consecutive reporting actions into a node group for that bed. Each reporting time point is a child node within the group. Next, perform the same operation for all beds in the hospital: create an independent node group for each bed, containing the time points of all reported events for that bed, sorted by time. Finally, these bed node groups are arranged in a certain order (e.g., alphabetical order of bed numbers) to form a hospital-wide "bed index." This index is a list, where each element is a bed, and each bed lists the time-series nodes of all reported events for that bed. For example, the bed index structure is as follows: {"12A-03": [Reporting time 1: 2024-04-06-09:26:30, Reporting time 2: 2024-04-07-15:10:00], "15B-01": [Reporting time 1: 2024-04-06-10:10:22, ...], ...}. This index provides a quick entry point for subsequently mapping responsibility trajectories to specific beds.

[0046] S512: Based on the bed index and responsibility point sequence, map the handover responsibility, shift handover responsibility, patrol responsibility, execution responsibility, abnormal handling responsibility and reporting responsibility to the corresponding bed nodes, mark the bedside responsibility turning point according to the responsibility handover break point, mark the handling implementation position according to the execution signature, abnormal recovery, alarm cancellation and reporting action position, and generate a dual-track position map by associating the two types of positions; Using the bed index as the entry point and combining it with the responsibility point sequence, for each bed in the bed index, such as bed "12A-03", all responsibility event sequences corresponding to that bed (i.e., the complete sequence generated in S213) are extracted from the responsibility point sequence. Then, a mapping operation is performed to attach each point in these responsibility event sequences to the node of that bed under the bed index. In this way, for bed "12A-03", it not only has the reporting time node, but also the time points of a series of responsibility events such as shift handover, shift takeover, rounds, execution, and abnormal handling. Next, two key marking actions are performed. The first marking action: marking the "bedside responsibility transition position". This position corresponds to the handover point of nursing responsibility. In the responsibility point sequence, the "cross-shift jump landing point" and the shift handover and takeover responsibility events before and after this point are identified. These points (such as shift handover check-out point, cross-shift jump point, and takeover confirmation point) are connected and marked as a trajectory, namely the "responsibility transition trajectory". For bed "12A-03", the responsibility transition point consists of the events: shift handover check-in (07:55:00) -> inter-shift transfer (08:00:00) -> shift handover confirmation (08:05:12). The second marking action is marking the "treatment implementation location". This location corresponds to the final execution and feedback point of the nursing action. In the responsibility point sequence, the event points "execution signature", "abnormal recovery", "alarm deactivated", and "reporting action" are identified. These points are connected and marked as another trajectory, namely the "treatment implementation trajectory". For bed "12A-03", the treatment implementation location consists of the events: execution signature (09:20:41) -> abnormal recovery (09:25:10) -> alarm deactivated (09:26:05) -> reporting action (09:26:30). After completing the double-marking, these two types of locations (responsibility transition trajectory and handling implementation trajectory) are linked together and plotted on the same two-dimensional graph with time as the horizontal axis and bed position as the vertical axis, generating a dual-track map. In this map, it can be clearly seen how the handover (transition) of responsibility and the execution (implementation) of responsibility unfold over time for the same bed position. These two trajectories may have temporal sequence and logical connection.

[0047] S513: Based on the dual-track bitmap, check the correspondence between the bedside responsibility turning point and the treatment implementation point and the bed, check the time sequence of each node and the continuity of responsibility labels, and synchronously write the bedside responsibility turning point trajectory and the treatment implementation trajectory into the nurse station terminal display page, associate the bed number, node position and trajectory segment identifier, and establish the overall trajectory analysis results of nursing events. Based on the dual-track floor plan, a final check and display were conducted. First, for each bed, the correspondence between the "responsibility transition trajectory" and the "treatment implementation trajectory" was checked. For bed "12A-03," the check revealed that the responsibility transition trajectory occurred around 08:00 AM, and the treatment implementation trajectory occurred after 09:20 AM. The two trajectories are sequential in time and both correspond to the same patient (same bed). This indicates that after completing the handover responsibility, the nurse successfully handled the patient's abnormal situation and completed the reporting in the subsequent time period. The correspondence is correct. Next, the time sequence of each node and the continuity of responsibility labels were checked. It was reconfirmed whether the event time within the responsibility transition trajectory increased (07:55:00 < 08:00:00 < 08:05:12) and whether the responsibility labels were continuous (handover -> jump -> takeover). Similarly, the time order within the treatment implementation trajectory is also ascending (09:20:41 < 09:25:10 < 09:26:05 < 09:26:30). Continuity verification passed. After verification, the generated dual-track bitmap data, including the two trajectories for each bed and all event nodes on them, will be pushed to the terminal display page at the nurse station in real time via the WebSocket interface. On the display page, a Gantt chart or timeline is used to display time horizontally and beds vertically. For bed "12A-03", two lines of different colors will be displayed in its row: one (e.g., blue) represents the responsibility transition trajectory, extending from 07:55:00 to 08:05:12; the other (e.g., red) represents the treatment implementation trajectory, extending from 09:20:41 to 09:26:30. When the mouse hovers over a node on the trajectory (e.g., "Execution Signature"), a pop-up will display specific information: event type, occurrence time, executing nurse, associated abnormal events, etc. Finally, the backend system associates and stores the bed number of each bed, the position of each node on its trajectory (e.g., 12A-03-1 represents the first trajectory point of that bed), and the identifier of the trajectory (e.g., "Responsibility Turnaround" or "Disposal Implementation") to form a complete "Overall Trajectory Analysis Result of Nursing Events". This result can be used by head nurses and quality control departments for retrospective analysis to assess the timeliness, continuity, and effectiveness of nursing work.

[0048] An artificial intelligence-based nursing event analysis system includes: The time-stamped acquisition module is used to perform S1: collect the handover confirmation time, medical order signing time, and execution signature time of nurses' handheld terminals in inpatient wards; collect the abnormal start time, abnormal recovery time, and alarm cancellation time of bed monitoring equipment; collect the nursing document recording time; rewrite the landing point of each record according to a unified timeline; and establish a nursing event time-stamped chain. The site indexing module is used to execute S2: based on the nursing event time-stamped chain, it arranges the order of handover check-out, handover confirmation, first round, doctor's order receipt, execution signature, abnormality initiation, abnormality recovery, and alarm event reporting, marks the responsibility gap landing point, cross-shift jump landing point, and key treatment actions, and outputs the responsibility site sequence; The semantic merging module is used to execute S3: based on the responsibility site sequence, it reads the ward verification case handover records and nursing document records, inputs the unified expression of the artificial intelligence semantic merging unit, rearranges the nursing case site order according to the same writing standard, and obtains nursing control benchmark data; The clue screening module is used to execute S4: Based on the responsibility site sequence and nursing control benchmark data, the first round of comparison is performed on the responsibility gap landing point, the cross-shift jump landing point, and the abnormal start time relative to the first inspection position. The second round of comparison is performed on the abnormal recovery time, alarm cancellation time, and the relationship between key treatment actions. Artificial intelligence is used to associate and merge consistent cases, and deviating cases are simultaneously eliminated to generate an intelligent mapping list. The analysis and aggregation module is used to execute S5: based on the intelligent mapping list, it reads the bed location corresponding to the reported action, maps the current nursing event bedside responsibility turning point trajectory and treatment implementation trajectory, and synchronously writes it into the nurse station terminal display page to generate the overall trajectory analysis results of the nursing event.

[0049] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A nursing event analysis method based on artificial intelligence, characterized in that, Includes the following steps: S1: Collect the handover confirmation time, doctor's order signing time, and execution signature time of nurses in the inpatient ward; collect the abnormal start time, abnormal recovery time, and alarm cancellation time of the bed monitoring equipment; collect the nursing document recording time; rewrite the landing point of each record according to a unified timeline; and establish a nursing event time-stamped chain. S2: Based on the nursing event time-stamped chain, arrange the order of handover check-out, handover confirmation, first round, doctor's order receipt, execution signature, abnormality initiation, abnormality recovery, and alarm event reporting, mark the responsibility gap, cross-shift jump point, and key treatment actions, and output the responsibility point sequence; S3: Based on the responsibility site sequence, read the ward verification case handover record and nursing document record, input the unified expression of the artificial intelligence semantic merging unit, rearrange the nursing case site order according to the same writing standard, and obtain nursing control benchmark data; S4: Based on the responsibility site sequence and the nursing control benchmark data, perform the first round of comparison for the responsibility gap landing point, the cross-shift jump landing point, and the abnormal start time relative to the first inspection position. Perform the second round of comparison for the abnormal recovery time, alarm cancellation time, and the relationship between key treatment actions. Use artificial intelligence to associate and merge consistent cases, simultaneously remove deviating cases, and generate an intelligent mapping list. S5: Based on the intelligent mapping list, read the bed location corresponding to the reported action, map the bedside responsibility turning point trajectory and treatment implementation trajectory of the current nursing event, and simultaneously write it into the nurse station terminal display page to generate the overall trajectory analysis result of the nursing event.

2. The nursing event analysis method based on artificial intelligence according to claim 1, characterized in that: The nursing event time-stamped chain includes time anchor markers, event sequence tags, and source attribution tags; the responsibility site sequence includes responsibility acceptance segment, intervention segment, and alarm connection segment; the nursing reference data includes semantic standard terms and case reference templates; the intelligent mapping list includes candidate case markers, trajectory matching tags, and deviation elimination markers; the overall trajectory analysis results of the nursing event in step five include responsibility evolution profile, intervention response profile, and nursing event judgment conclusion.

3. The nursing event analysis method based on artificial intelligence according to claim 1, characterized in that, The specific steps for obtaining the nursing event time-stamped chain are as follows: S111: Collect the handover confirmation time, doctor's order signing time, and execution signature time of nurses in the inpatient ward using their handheld terminals, verify the time field format, establish a correspondence between the bed number and the doctor's order number, and generate a nursing operation time mapping table. S112: Obtain the abnormal start time, abnormal recovery time, alarm cancellation time and nursing record time of the bed monitoring equipment, compare the order of each time according to the nursing operation time mapping table, and establish an abnormal handling time sequence table; S113: Based on the abnormal handling time sequence table, associate it with the nursing operation time mapping table, rewrite the time points of handover confirmation, medical order receipt, execution signature, abnormal start, abnormal recovery, alarm cancellation, and document record to a unified time axis, and establish a nursing event time stamp chain according to bed number and event sequence.

4. The nursing event analysis method based on artificial intelligence according to claim 1, characterized in that, The specific steps for obtaining the responsible site sequence are as follows: S211: Based on the nursing event time-stamped chain, extract the landing points of handover check-out, handover confirmation, first round, medical order receipt, execution signature, abnormal start, abnormal recovery, and alarm event reporting, check the order of the events along a unified time axis, and if the landing points are the same second, rearrange the order according to the rule that nursing operation precedes abnormal state and abnormal state precedes reporting record, and establish a time sequence point ranking list. S212: Obtain the preset shift time schedule, combine it with the time sequence location list, compare the interval between handover check-out and handover confirmation, the interval between handover confirmation and first round, and the interval between medical order receipt and execution signature, determine the responsibility gap landing point and cross-shift jump landing point based on the shift boundary time, mark the execution signature between the abnormal start and abnormal recovery as the key handling action, and generate a responsibility mark index. S213: Based on the responsibility marker index, associate the time sequence point list, write the responsibility gap landing point, cross-shift jump landing point, and key handling action into the corresponding event position, and connect the handover responsibility, shift takeover responsibility, patrol responsibility, execution responsibility, abnormal handling responsibility and reporting responsibility in chronological order to establish a responsibility point sequence.

5. The nursing event analysis method based on artificial intelligence according to claim 1, characterized in that, The specific steps for obtaining the nursing baseline data are as follows: S311: Based on the responsibility site sequence, read the locally stored desensitized ward case handover records and nursing document records, check the correspondence by bed number, shift number, and event tag, split the handover statement and document statement, check the consistency of record time and responsibility position, check the consistency of statement source field, eliminate duplicate points, aggregate statements of the same bed and shift into continuous segments, and establish a case statement matching set; S312: Input the case statement arrangement set into the artificial intelligence semantic merging unit, retrieve synonyms, abbreviations, inverted statements and tense differences between handover statements and document statements, merge statements pointing to the same responsible event into unified event tags, rewrite the description time point into a unified time term, rewrite the disposal verb into a unified action word, rewrite the abnormal state into a unified state word, correct the subject reference relationship of the statement, and generate a writing caliber mapping table; S313: Based on the writing caliber mapping table, associate the case statement arrangement set with the responsibility site sequence, rewrite the nursing case site name according to the unified event label, rearrange the site order according to the order of handover responsibility, takeover responsibility, inspection responsibility, execution responsibility, abnormal handling responsibility and reporting responsibility, check the correspondence between the rearranged sites and responsibility positions, calculate the consistency index of site responsibility, map the handover statement sequence and the document statement sequence into a coaxial comparison structure, and establish nursing comparison benchmark data.

6. The nursing event analysis method based on artificial intelligence according to claim 5, characterized in that, The formula for calculating the consistency index corresponding to the site responsibility is as follows: ； in, The consistency index corresponds to the responsibility of the representative site. The total number of liability events in representative cases, The indices representing the responsible events range from 1 to N. The statement representing the nth responsible event maps to confidence. This represents the position of the nth responsible event after rearrangement. This represents the standard liability ranking of the nth liability event. This represents the actual time interval between the nth and (n+1)th responsibility events. This represents the standard time interval between the nth and (n+1)th liability events. This represents a preset minimum positive number used to avoid division by zero.

7. The nursing event analysis method based on artificial intelligence according to claim 1, characterized in that, The specific steps for obtaining the smart mapping list are as follows: S411: Based on the responsibility site sequence and the nursing control baseline data, align the sites according to bed number, case number, and responsibility label, verify whether the responsibility gap falls within the baseline gap range, check whether the shift markers before and after the cross-shift jump point are continuous, compare whether the abnormal start time and the order of the first round of inspections are consistent, mark consistent items as coaxial items, mark offset items as difference items, and generate the first round of site marker clusters; S412: Based on the first round of site marker clusters, associate the nursing control baseline data, check whether the order of abnormal recovery time and alarm cancellation time corresponds, verify whether the key treatment action is located between the abnormality start and abnormal recovery, compare whether the relationship between the key treatment action and the alarm cancellation time is consistent, write the continuous relationship into the association label, write the reverse order relationship into the deviation label, and establish the second round of relationship screening clusters. S413: Based on the first round of site marker clusters and the second round of relationship filtering clusters, input the artificial intelligence association and merging unit to aggregate cases with the same bed, the same responsibility label, and the same sequence relationship. Remove cases with gap offset, shift break, reverse order restoration, and reverse order removal from the candidate set. Establish a mapping relationship between case number and responsibility label to generate an intelligent mapping list.

8. The nursing event analysis method based on artificial intelligence according to claim 1, characterized in that, The specific steps for obtaining the overall trajectory analysis results of the nursing events are as follows: S511: Based on the intelligent mapping list, read the bed location corresponding to the reported action, verify the correspondence between the bed number, case number, responsibility label and reporting time, merge consecutive reported actions of the same bed into the same group of nodes, rearrange the bed node positions according to time order, and establish a bed index. S512: Based on the bed index and the responsibility point sequence, map the handover responsibility, shift handover responsibility, patrol responsibility, execution responsibility, abnormal handling responsibility and reporting responsibility to the corresponding bed nodes, mark the bedside responsibility turning point according to the responsibility handover breakpoint, mark the handling implementation position according to the execution signature, abnormal recovery, alarm cancellation and reporting action position, and generate a dual-track bitmap by associating the two types of positions; S513: Based on the aforementioned dual-track bitmap, check the correspondence between the bedside responsibility turning point and the treatment implementation point and the bed, verify the time sequence of each node and the continuity of responsibility labels, synchronously write the bedside responsibility turning point trajectory and the treatment implementation trajectory into the nurse station terminal display page, associate the bed number, node position and trajectory segment identifier, and establish the overall trajectory analysis results of the nursing event.

9. A nursing event analysis system based on artificial intelligence, characterized in that, The system is used to implement the method according to any one of claims 1-8, comprising: The time stamp acquisition module collects the handover confirmation time, medical order signing time, and execution signature time of nurses' handheld terminals in inpatient wards; it also collects the abnormal start time, abnormal recovery time, and alarm cancellation time of bed monitoring equipment; and it collects the nursing document recording time. It rewrites the landing point of each record according to a unified timeline and establishes a nursing event time stamp chain. The site indexing module, based on the nursing event time stamp chain, arranges the sequential positions of handover check-out, shift acceptance confirmation, first round of inspection, medical order receipt, execution signature, abnormality initiation, abnormality recovery, and alarm event reporting, marks the responsibility gap landing point, cross-shift jump landing point, and key treatment actions, and outputs the responsibility site sequence. The semantic compilation module reads the ward verification case handover records and nursing document records according to the responsibility site sequence, inputs them into the artificial intelligence semantic merging unit for unified expression, rearranges the nursing case site order according to the same writing standard, and obtains nursing control benchmark data. The clue screening module, based on the responsibility site sequence and the nursing control benchmark data, performs the first round of comparison for the responsibility gap landing point, the cross-shift jump landing point, and the abnormal start time relative to the first inspection location. It performs the second round of comparison for the abnormal recovery time, alarm cancellation time, and the relationship between key treatment actions. It uses artificial intelligence to associate and merge consistent cases, simultaneously eliminates deviating cases, and generates an intelligent mapping list. The analysis and aggregation module reads the bed location corresponding to the reported action based on the intelligent mapping list, maps the bedside responsibility turning point trajectory and treatment implementation trajectory of the current nursing event, and simultaneously writes it into the nurse station terminal display page to generate the overall trajectory analysis result of the nursing event.

Images