Intelligent checking, error early warning and traceability management method for ICU nursing shift information

Abstract

The application relates to the field of medical information processing and discloses an ICU nursing shift information intelligent checking, error early warning and trace management method, which comprises the following steps: acquiring monitor and life support equipment data, generating time axis aligned discrete time series based on global time resampling; differentially detecting a jump point of a positioning instruction, verifying physical work integral to generate an actual intervention node, and calculating a physiological response transition time window; extracting a shift table form to generate a subjective declaration set and a subjective declaration final state vector; comparing time deviation and parameter deviation of the subjective and objective data, and triggering an early warning when the deviation exceeds a threshold; and if a shift time falls within the physiological response transition time window, freezing the shift table form signature authority. By converting subjective document review into objective checking based on equipment bottom layer data, the application excludes artificial recording errors and omissions, quantifies a drug effect physical lag process, and effectively prevents treatment interruption risks caused by consumable depletion and physical sign fluctuation during a shift period.

Description

Technical Field

[0001] This invention relates to the field of medical information processing, specifically to a method for intelligent verification, error warning, and traceability management of ICU nursing handover information. Background Technology

[0002] Nursing handover in the intensive care unit (ICU) is a crucial step in ensuring the continuity of patient care. Currently, ICUs are typically equipped with multi-parameter monitors and various life support devices such as infusion pumps and ventilators. During the handover process, nursing staff need to read the operating parameters of these medical devices and the patient's vital signs data, and enter the progress of medical orders, equipment parameter settings, and remaining consumables into an electronic form system to generate a handover document. The incoming staff then use this document to conduct bedside handover and verify equipment status to ensure a smooth transition of all medical interventions.

[0003] The existing shift handover verification process mainly relies on the incoming staff reading the written records and comparing them with the on-site equipment displays. Because the local clocks of the heterogeneous medical equipment in the intensive care unit operate independently, lacking a unified global time reference, there is a drift and misalignment between the underlying time-series data received by the system and the intervention time points in the nursing documentation. This time-dimension discrepancy makes it difficult for the information system to objectively align the actual physical work feedback from the equipment with the subjective operational statements entered manually. This results in the system being unable to conduct substantive verification of the accuracy of the written content, posing a risk of information errors or omissions, and making it difficult to trigger automated error warnings during the shift handover process.

[0004] Current shift handover management lacks dynamic quantitative tracking of treatment intervention delays and consumable consumption status. When operating infusion equipment, due to the dead space volume and set flow rate of the infusion tubing, there is a certain physical transition period between the delivery of medication from the equipment and the resulting physiological changes in the patient. If the shift handover occurs during this transition period, before the patient's vital signs have reached a steady state, routine handover confirmation procedures can easily mask potential risks of fluctuating patient conditions. Furthermore, current handover procedures for medications and other consumables rely heavily on manual observation and estimation, failing to accurately calculate remaining supplies and predict depletion times based on actual equipment workload. This hinders cross-shift consumable risk control and can easily lead to treatment interruptions due to consumable depletion shortly after the shift handover. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent verification, error warning, and traceability management method for ICU nursing handover information. This method solves the problem that existing handover processes in intensive care units rely on manual recording and subjective review, and that the local clocks of heterogeneous medical devices drift and become misaligned, resulting in a lack of verification between the physical operating status of the devices and nursing documentation.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of this invention provides a method for intelligent verification, error early warning, and traceability management of ICU nursing handover information, comprising the following steps: S10: Obtain the vital sign response vector output by the multi-parameter monitor and the digital command vector and physical feedback vector output by the life support device; perform unified frequency resampling based on the global time base to generate a time-axis aligned discrete time series. S20, call the equipment capability profile matrix to perform differential detection on the instruction vector in the discrete time series to locate the instruction jump point, construct a dynamic confirmation integral window within the time span corresponding to the instruction jump point to verify the actual physical work feedback integral, generate actual intervention nodes after confirmation and construct an objective intervention set, and calculate the physiological response transition time window based on the equipment type and anchor it to the corresponding actual intervention node. S30: Capture the electronic form of the shift handover at the end of the shift handover observation time window, extract the text and structured data and assign semantic timestamps to generate a set of subjective statements, and at the same time extract the final operating parameters of the equipment in the form to generate a final state vector of subjective statements. S40: Calculate the time deviation distance between the associated nodes in the objective intervention set and the subjective declaration set, and trigger a terminal blocking warning when the time deviation distance exceeds the time deviation tolerance threshold; compare the objective final state vector and the subjective declaration final state vector corresponding to the discrete time series, and trigger a final state conflict warning when a state deviation occurs. S50: When the end time of the handover observation time window is determined to fall within the physiological response transition time window corresponding to any actual intervention node, a cross-shift intervention connection identifier is generated in the handover electronic form, the confirmation signature permission of the current handover vital signs baseline is frozen, and the handover signature link is restored after the relevant physiological response transition time window is exhausted.

[0007] This invention uses underlying data time alignment and mechanical feedback verification to perform differential comparison between the actual operating parameters of the execution equipment and the handover document records; combined with the dead cavity volume of the infusion pipeline and the set flow rate to quantify the drug efficacy delay period, it freezes the handover signature during the physiological response transition period, and triggers a risk control locking mechanism based on the remaining amount of consumables, thereby preventing the risk of information omission and treatment interruption in the handover process.

[0008] Furthermore, the steps for generating time-aligned discrete time series specifically include: the system obtains standard system time through the network time protocol to establish a global time reference; when receiving data frames from devices, the original timestamp is stripped, and a global arrival timestamp is added based on the global time reference and stored in the buffer queue; a discrete time node sequence is generated based on the preset reference sampling period, linear interpolation resampling is performed on the symptom response vector, and zero-order hold resampling is performed on the digital command vector and physical feedback vector.

[0009] Furthermore, the steps of performing differential detection to locate the command transition point specifically include: reading the minimum response step size and mechanical action delay parameters defined in the device capability profile matrix; calculating the command difference between adjacent discrete time nodes; and when the command difference is greater than the minimum response step size and the duration of the continuous non-zero state meets the set duration condition, the starting time of the state change is located as the command transition point.

[0010] Furthermore, the steps for verifying the actual physical work feedback integral specifically include: setting the length of the dynamic verification integration window according to the equipment type; accumulating the value of the digital command vector within the dynamic verification integration window by multiplying it by the reference sampling period to calculate the theoretical total work; extracting the increment of the physical feedback vector within the same dynamic verification integration window as the actual physical work feedback integral; and determining whether the ratio of the absolute value of the difference between the theoretical total work and the actual physical work feedback integral to the theoretical total work falls within the tolerance error band condition. If it is satisfied, the physical work verification is deemed successful.

[0011] Furthermore, the steps for calculating the physiological response transition time window based on device type specifically include: for infusion-type life support devices, extracting the dead space volume parameter of the infusion tubing, dividing the dead space volume parameter by the set flow velocity vector, calculating the drug infusion hysteresis period, and using the drug infusion hysteresis period as the physiological response transition time window; for respiratory support-type life support devices, determining the physiological response observation time window based on the waveform characteristics or alarm status cycle after parameter changes, and classifying them as the physiological response transition time window.

[0012] Furthermore, the method also includes: configuring a maximum protection threshold for the physiological response transition time window, and forcibly truncating to the maximum protection threshold when the calculated period or time window exceeds the maximum protection threshold.

[0013] Furthermore, the steps of extracting text and structured data and assigning semantic timestamps specifically include: calling the built-in clinical standard dictionary and regular expression matching engine to identify nursing operations and medical order execution events in unstructured text; extracting time strings from the text, directly converting time strings in absolute time form into standard timestamps; and subtracting the relative offset duration from the form system submission time to obtain the semantic timestamp for time strings in relative time description.

[0014] Furthermore, the steps for calculating the time deviation distance of associated nodes specifically include: traversing the set of subjective statements, searching for clinical statement nodes that match the attributes of the actual intervention node, where matching attributes include at least device identifier matching and event type matching; when there are multiple candidate clinical statement nodes, selecting the clinical statement node with the smallest time distance as the matching result, and calculating the time deviation distance by subtracting the semantic timestamp of the subjective record from the time when the underlying objective intervention occurred.

[0015] Furthermore, the steps to trigger the terminal blocking warning specifically include: locking the submission process of the electronic handover form on the front-end interface of the nursing workstation, retrieving and displaying the device waveform data and vital sign time-series slice data before and after the actual intervention point, and prompting the correction of the document record.

[0016] Furthermore, the steps for comparing the objective final state vector with the subjective declared final state vector specifically include: performing index alignment on the objective final state vector and the subjective declared final state vector based on the device identification code; filling in missing state markers to generate state missing conflict records for positions where the device identification codes are not equal; performing numerical difference calculation on the aligned positions, and sending a misalignment alarm to the mobile terminal to prompt on-site verification when the deviation component exceeds the device's allowable error.

[0017] Furthermore, the method also includes a closed-loop traceability step for consumables: extracting the initial nominal volume of the drug solution package in the tubing of infusion-type life support equipment; accumulating the actual physical work feedback integral to obtain the cumulative total consumption; subtracting the cumulative total consumption from the initial nominal volume to obtain the physical balance of the drug solution at the time of shift handover; and transmitting the physical balance to the medical supplies verification database for storage.

[0018] Furthermore, the method also includes the step of triggering a cross-shift dynamic risk control lock: dividing the physical surplus by the current set flow rate of the infusion-type life support equipment to calculate the expected depletion time; when the expected depletion time is less than the system-set safe continuity threshold, the cross-shift dynamic risk control lock is triggered.

[0019] Furthermore, when infusion-type life support equipment has a remote control interface, the execution logic of the cross-shift dynamic risk control lock is as follows: the edge computing gateway sends a permission contraction command to the equipment, and the operation permission to increase the flow rate is locked on the control panel of the equipment itself.

[0020] Furthermore, when infusion-type life support equipment does not have a remote control interface, the execution logic of the cross-shift dynamic risk control lock is as follows: a forced intervention interface pops up at the front end of the nursing workstation, requiring the outgoing and incoming nurses to scan the code at the bedside to confirm the standby medication before unlocking the form terminal.

[0021] Furthermore, the method also includes: structuring and assembling the actual intervention nodes, subjective declaration sets, and state deviation records, attaching absolute timestamps, and generating closed-loop traceability logs by calculating summary feature values ​​through a hash algorithm.

[0022] The second aspect of the present invention provides an intelligent verification, error warning and traceability management system for ICU nursing handover information, used to execute the method provided in the first aspect of the present invention. The system includes a synchronous sampling module, a node verification module, a declaration parsing module, a differential comparison and early warning module and a traceability risk control module. The functional programs corresponding to each module run collaboratively in the edge computing gateway and the clinical information system.

[0023] Furthermore, the edge computing gateway has a system capability profile matrix index table deployed inside, which is used to map the communication delay compensation parameters and allowable tolerance error band limits of different types of bedside devices.

[0024] This invention provides a method for intelligent verification, error early warning, and traceability management of ICU nursing handover information. It has the following beneficial effects:

[0025] 1. This invention resamples data output from monitors and life support devices at a unified frequency based on a global time reference. It then uses digital commands and physical feedback integrals to generate objective actual intervention points, which are then compared with subjective declaration data extracted from electronic shift handover forms to calculate discrepancies. This mechanism solves the verification misalignment problem caused by clock asynchrony in heterogeneous medical devices, transforming the manual document review during shift handover into objective data verification based on the actual physical work done by the equipment, thus eliminating the risk of errors and omissions caused by manual recording.

[0026] 2. This invention calculates the drug infusion lag period and establishes a physiological response transition time window by extracting the dead space volume of the infusion tubing and the set flow rate of the equipment. When the handover time falls within this time window, the confirmation signature permission of the handover form is forcibly frozen. This setting quantifies the physical lag process between drug delivery and the patient's physiological response, preventing the risks associated with handover when the patient's vital signs are not yet stable or the drug effect has not yet appeared, and ensuring a smooth transition of treatment interventions across shifts.

[0027] 3. This invention objectively calculates the total cumulative consumption and remaining physical quantity of consumables by accumulating the feedback integral of the actual physical work done by the life support equipment. Combined with the set flow rate at shift handover, it calculates the estimated depletion time. When the estimated depletion time is lower than the safe continuation threshold, a dynamic risk control mechanism is triggered, restricting the equipment's ability to increase the flow rate. This process establishes a digital tracking link for consumable reserves, preventing treatment interruptions caused by the infusion solution running out shortly after shift handover and not being detected in time by the incoming staff. Attached Figure Description

[0028] Figure 1 This is a system framework diagram of the present invention; Figure 2 This is a flowchart of the method of the present invention; Figure 3 This is a schematic diagram illustrating the discrete time series generation and resampling principle of the present invention. Figure 4 This is the logic diagram for physical work verification and physiological hysteresis anchoring of the present invention. Figure 5 This is a flowchart of the semantic parsing and subjective statement extraction process for the handover documents of the present invention; Figure 6 This is a logic diagram of the cross-dimensional asymmetric difference comparison and early warning distribution of the present invention; Figure 7 This is a flowchart of the consumable closed-loop traceability and cross-shift dynamic risk control lock of the present invention; Figure 8 This is a high-dimensional feature projection map of the multi-source heterogeneous time-series data joint calibration and state verification of the present invention; Figure 9 This is a scatter plot showing the multidimensional protection effectiveness and error dimensionality reduction of the present invention.

[0029] Among them, 10. Synchronous sampling module; 11. Multi-parameter monitor; 12. Life support equipment; 13. Edge computing gateway; 14. Clinical information system; 20. Node verification module; 30. Declaration parsing module; 40. Differential comparison and early warning module; 50. Traceability risk control module. Detailed Implementation

[0030] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] See attached document Figure 1 This invention provides a hardware deployment environment for an intelligent verification, error early warning, and traceability management system for ICU nursing shift handover information.

[0032] The system is deployed in the local area network environment of the intensive care unit. The multi-parameter monitor 11 and life support equipment 12 are configured in the patient's bedside environment. The life support equipment 12 includes micro-infusion pumps and ventilators. For infusion-type life support equipment such as micro-infusion pumps, the system focuses on collecting their flow rate setting parameters, cumulative infusion volume and related feedback parameters, and performs drug efficacy hysteresis analysis based on the infusion tubing parameters.

[0033] For life support devices such as ventilators, the system focuses on collecting operating parameters such as ventilation mode, pressure parameters, tidal volume, flow rate, or alarm status, and performs node verification and status consistency check based on the feedback parameters corresponding to the respiratory support device.

[0034] The multi-parameter monitor 11 and life support device 12 are each configured with independent data communication interfaces. The edge computing gateway 13 establishes a low-level hardware connection with the multi-parameter monitor 11 and life support device 12 through the data communication interfaces. The edge computing gateway 13 establishes a communication connection with the clinical information system 14 through an internal local area network. The clinical information system 14 includes a central data server at the hospital backend and nursing terminal interaction devices deployed in the wards.

[0035] The system includes a synchronous sampling module 10, a node verification module 20, a declaration parsing module 30, a differential comparison and early warning module 40, and a traceability risk control module 50.

[0036] See attached document Figure 2 This invention provides a method for intelligent verification, error early warning, and traceability management of ICU nursing handover information, comprising the following steps: S10, the synchronous sampling module 10 and the main server of the clinical information system 14 execute the network time protocol to synchronously establish a global time reference; the synchronous sampling module 10 collects the vital sign response vector output by the multi-parameter monitor 11 in real time, and synchronously collects the digital command vector and physical feedback vector output by the life support device 12; the synchronous sampling module 10 performs uniform frequency resampling on the vital sign response vector, digital command vector and physical feedback vector according to the global time reference, and outputs a discrete time series aligned on the same time axis; S20, the node verification module 20 acquires the aligned discrete time series, reads the device capability profile matrix, performs first-order differential detection on the digital command vector, and locates the command jump point that meets the duration condition; the node verification module 20 constructs a dynamic verification integral window within the time span corresponding to the command jump point, calculates the theoretical total work done by the digital command vector within the dynamic verification integral window, and extracts the actual physical work feedback integral of the physical feedback vector within the same dynamic verification integral window; the node verification module 20 compares the theoretical total work done with the actual physical work feedback integral, and when the tolerance error band condition is met, encapsulates the parameter difference before and after the command jump point and the corresponding vital sign change characteristic value as an actual intervention node, generating an objective intervention set; For infusion-type life support devices, the node verification module 20 extracts the dead space volume parameter of its infusion tubing and calculates the drug infusion lag period when the drug reaches the patient by combining the set flow velocity vector corresponding to the command jump point. For respiratory support-type life support devices, the node verification module 20 determines the corresponding physiological response observation time window based on the feedback parameters, waveform characteristics or alarm status after the device parameters change.

[0037] The node confirmation module 20 unifies the aforementioned drug infusion delay period or physiological response observation time window into a physiological response transition time window, and anchors the physiological response transition time window as an entity attribute to the actual intervention node; the physiological response transition time window can correspond to the drug infusion delay process caused by infusion-type life support equipment, or to the physiological response observation process after the parameter adjustment of respiratory support-type life support equipment. S30, the declaration parsing module 30 captures the handover electronic form data in the clinical information system 14 at the end of the handover observation time window; the declaration parsing module 30 parses the structured checkboxes and text fields in the handover electronic form, extracts nursing operations and medical order execution events and assigns semantic timestamps, generates declaration clinical nodes and constructs a set of subjective declarations; the declaration parsing module 30 extracts the final operating parameters of the equipment in the handover electronic form and generates a final state vector of subjective declarations; S40, the differential comparison and early warning module 40 acquires the objective intervention set and the subjective statement set, and calculates the time deviation distance of the associated nodes; when the time deviation distance exceeds the handover time sequence deviation tolerance threshold, the differential comparison and early warning module 40 sends the equipment instruction change data and the vital sign time sequence slice data to the clinical information system 14, and triggers the omission blocking alarm on the nursing terminal interface. Among them, the vital signs time series slice data takes the objective occurrence time corresponding to the actual intervention node as the center, and extracts the equipment parameter change data and vital signs data within a preset time range before and after it, which are used for traceability display on the nursing terminal interface; the differential comparison and early warning module 40 calculates the state deviation results between the objective final state vector extracted from the discrete time series cached by the edge computing gateway 13 at the end of the shift handover observation time window and the subjectively declared final state vector, and triggers a final state conflict early warning when the calculated component exceeds the tolerance error band; S50, the traceability risk control module 50 accumulates the actual physical work feedback integral to generate infusion consumable consumption data, and transmits the infusion consumable consumption data to the material verification database of the clinical information system 14 for storage and recording; when the traceability risk control module 50 identifies that the end time of the shift handover observation time window falls within the physiological response transition time window anchored by the actual intervention node, or when the end time of the shift handover observation time window falls within any of the physiological response transition time windows corresponding to multiple actual intervention nodes, the traceability risk control module 50 forcibly generates a cross-shift intervention connection identifier in the shift handover electronic form and injects the delayed vital sign observation task into the subjective declaration final state vector.

[0038] The handover signature permission is frozen through the nursing terminal interface until the relevant physiological response transition time window is exhausted, or after manual review, the handover signature link is restored.

[0039] The following section will elaborate on each step and underlying computational logic of the above method flow in conjunction with specific implementation methods.

[0040] See attached document Figure 3 In clinical intensive care settings, bedside devices of different brands and models have their own independent local crystal oscillators. Long-term operation causes clock drift, leading to misalignment of the data timeline. To provide a unified data computing platform, the intelligent verification, error warning, and traceability management method for ICU nursing handover information provided by this invention includes the following sub-steps for data acquisition and time alignment of heterogeneous devices.

[0041] S101, in this embodiment, the synchronization sampling module 10 establishes a global time reference for the system. Specifically, the synchronization sampling module 10 sends a time synchronization request to the main server of the clinical information system 14 through the network port of the edge computing gateway 13. The main server distributes the standard system time to the edge computing gateway 13. The synchronization sampling module 10 maintains the standard system time in the memory of the edge computing gateway 13 and establishes it as the global time reference. To limit the scope of business analysis, the system uses the pre-set management schedule of medical institutions as a global time base. The observation time window for the current shift is captured and marked as follows. .symbol Indicates the time of shift handover sign-in, symbol This indicates the handover confirmation time. For the underlying data packet interaction and delay compensation of Network Time Protocol (NTP) synchronization, those skilled in the art can implement it using standard NTP stack programs. Its communication handshake and offset calculation mechanisms are well-known technologies in the field and will not be elaborated upon here.

[0042] S102, the synchronous sampling module 10 performs raw data acquisition and forced timestamp overwriting. The multi-parameter monitor 11 and life support equipment 12 in the intensive care unit continuously generate data frames during operation. Since the devices are not connected to a unified clock source, their generated raw timestamps suffer from clock drift and reference misalignment. As a preferred method, the synchronous sampling module 10 pulls data frames in parallel through the physical communication interface of the edge computing gateway 13. Upon receiving a data frame, the synchronous sampling module 10 strips the original timestamps from the devices within the data frame and reads the global time reference currently maintained by the system. Attach a global arrival timestamp to the data frame .

[0043] In one implementation, when the device simultaneously outputs the original event time, frame number, or auxiliary fields that can be used to estimate communication delay, the synchronization sampling module 10 can correct the global arrival timestamp by combining the communication delay estimation results; when the device does not have the aforementioned fields, the global arrival timestamp is used as the approximate occurrence time under a unified time base.

[0044] The synchronous sampling module 10 parses the data frame according to the field definitions of the device communication protocol. For the data frame of the multi-parameter monitor 11, the synchronous sampling module 10 extracts physiological monitoring values ​​to form a vital sign response vector. For the data frames from the life support device 12, the synchronous sampling module 10 extracts the operating parameters set on the control panel to form a digital instruction vector. Synchronously extract the actual cumulative consumption value generated by the underlying motor sensor to form a physical feedback vector. The symptom response vector, digital command vector, and physical feedback vector all carry a unified global arrival timestamp. It is stored in the cache queue of the edge computing gateway 13 to support discretization processing.

[0045] In one implementation, the multi-parameter monitor 11 and the life support device 12 are connected to the edge computing gateway 13 via a serial port, Ethernet interface, or vendor-open communication interface. The edge computing gateway 13 is configured with a protocol adaptation layer to uniformly map data fields output by devices of different brands and models. For devices that only support data reading, the system performs data acquisition, verification, and early warning functions; for devices that also support the issuance of authorized control commands, the system further performs permission contraction or risk control functions.

[0046] S103, the synchronous sampling module 10 executes a unified frequency resampling algorithm for heterogeneous data. The push frequencies of devices from different manufacturers differ, causing the data in the buffer queue to exhibit an unequal distribution. The synchronous sampling module 10 sets the system's baseline sampling period. In practical applications, the reference sampling period The value is matched based on the highest output frequency of the multi-parameter monitor 11, and the value range is set between 0.5 seconds and 2 seconds to balance the system's computing power consumption and timing recording accuracy. The synchronous sampling module 10 is based on the reference sampling period. Generate discrete time node sequence , where the symbol It is a positive integer.

[0047] Synchronous sampling module 10 pairs of vital sign response vectors Perform linear interpolation resampling. Vital data are represented as continuously changing physical quantities; the synchronous sampling module 10 searches for discrete time nodes. Adjacent preceding arrival times With subsequent arrival time The resampled feature vector is calculated using the following formula. : ; In the formula, This represents the arrival time of the vital sign response vector. The actual collected values, This represents the arrival time of the vital sign response vector. The actual collected values. To avoid data out-of-bounds dead zones during real-time system processing, if no data has been received in the current buffer queue... For each data point, the synchronous sampling module 10 suspends the processing for a preset maximum tolerable delay time. If no new data is acquired within this timeout period, then... It is used as a substitute value for the current moment in the calculation.

[0048] Synchronous sampling module 10 pairs of digital command vectors and physical feedback vector Zero-order hold resampling is performed. Digital commands and physical consumption data exhibit state-maintaining characteristics; their operational state remains in a hold phase until two consecutive parameter change commands are issued. Synchronous sampling module 10 locates the distance to the discrete time node. The most recent and less than or equal to Arrival time The resampling command vector is calculated using the following formula. : ; In the formula, Indicates the arrival time of the digital command vector. The parameter status, This indicates the arrival time of the next state change instruction. In this operational logic, if the discrete time node... In the initial period when the device is powered on but has not yet received the first valid instruction, the module is assigned a shutdown safety state constant defined in the device communication protocol.

[0049] The zero-order hold resampling calculation method of the physical feedback vector is consistent with that of the digital instruction vector, and the synchronous sampling module 10 generates the resampling feedback vector accordingly. The synchronous sampling module 10 aggregates and summarizes the calculated resampled vital sign vector, resampled command vector, and resampled feedback vector, outputting a discrete time series aligned on the same time axis. This discrete time series provides underlying data support for the system's subsequent identification of real intervention nodes and execution differential comparison.

[0050] See attached document Figure 4 , Figure 4 This is a logic diagram for physical work verification and physiological hysteresis anchoring according to an embodiment of the present invention. After obtaining the underlying data aligned to the time axis, the system performs physical quantification on changes in device parameters to eliminate false changes caused by idling or operational errors. In this embodiment, the physical work verification and physiological hysteresis anchoring steps specifically include the following sub-step flow process.

[0051] S201, the node verification module 20 obtains the aligned discrete time series and calls the pre-stored device capability profile matrix. The device capability profile matrix records the inherent hardware attributes of various intensive care equipment, including minimum response step size and mechanical action delay parameters.

[0052] Preferably, the equipment capability profile matrix is ​​established based on one or more of the following: equipment model, factory parameters, on-site calibration results, or manual maintenance configuration, and stored in the edge computing gateway 13 or the clinical information system 14, for the node verification module 20 to index and retrieve according to the equipment identifier. In order to identify valid equipment control commands, the node verification module 20 reads the equipment capability profile matrix, performs first-order difference detection on the resampled command vector in the discrete time series, and calculates the command difference between adjacent discrete time nodes.

[0053] To eliminate noise interference generated when medical staff operate the equipment panel, the node verification module 20 determines whether the command difference is greater than the minimum response step size defined in the equipment capability profile matrix, and performs time window filtering accordingly. When the duration for which the command difference remains non-zero meets the set duration condition, the node verification module 20 identifies the start time of the state change as the command transition point. As a preferred method, the duration condition is set to 3 to 5 seconds, with the specific value dynamically adjusted according to the mechanical action delay parameters of the equipment.

[0054] S202, after the location command jump point, to prevent equipment idling due to pipeline blockage or sensor failure, the node verification module 20 verifies whether the underlying motor of the equipment has output physical work. The principle is to cross-verify the theoretical output or theoretical state change corresponding to the parameter setting change with the underlying feedback parameters. The node verification module 20 constructs a sequence of lengths forward within the time span corresponding to the command jump point. The dynamic confirmation integral window. Integral window length. Set according to the device type. For infusion-type life support devices, the integration window length is... The integral window length is set based on the stable execution cycle after a single standard flush, a single pump inlet, or parameter adjustments; for respiratory support life support equipment, the integral window length is... The feedback stabilization period, waveform change period, or alarm status refresh period are set based on the adjusted ventilation parameters. Preferably, the aforementioned integral window length... The value range can be set to 5 to 10 reference sampling periods. The node verification module 20 calculates the theoretical total work done by the digital command vector within the dynamic verification integration window using the following formula. : ; In the formula, This indicates the discrete time node where the instruction jump point occurs. Indicates at discrete time nodes The resampling instruction vector, i.e., the set operating parameters, Indicates the reference sampling period.

[0055] Simultaneously, the node verification module 20 extracts the increment of the resampled feedback vector in the discrete time series within the same dynamic verification integration window, and uses it as the actual physical work feedback integral. If some devices lack underlying feedback sensors, resulting in the inability to acquire resampled feedback vectors, the node verification module 20 initiates a degradation strategy. It marks the corresponding command transition point as a low-confidence intervention candidate node and uses the theoretical total work as a reference estimate for subsequent analysis, but does not directly determine whether the physical work verification has passed based on this estimate. The node verification module 20 compares the theoretical total work... Feedback integral with actual physical work done .

[0056] The node verification module 20 calculates the ratio of the absolute value of the difference between the two values ​​to the theoretical total work done. When the ratio falls within the tolerance error band, the physical work done is deemed verified. The upper limit of the tolerance error band is set according to the factory calibration accuracy of the device's sensors, ranging from 2% to 5%. After verification, the node verification module 20 encapsulates the parameter difference before and after the command jump point, as well as the characteristic values ​​of the vital signs change within the corresponding time period, into an independent data structure, defining it as an actual intervention node. For devices without underlying feedback sensors, the node verification module 20 encapsulates the corresponding node as a low-confidence actual intervention node or a node awaiting manual verification, and adopts a prompting strategy or relaxes the tolerance threshold in subsequent comparison and early warning stages. The node verification module 20 aggregates all actual intervention nodes within the observation time window to generate an objective intervention set for subsequent comparison.

[0057] S203, as a preferred method, for infusion-type life support devices, considering the physical spatial span of clinical intravenous infusion drugs, the node verification module 20 performs physiological-dimensional temporal correction on the actual intervention node. The node verification module 20 initiates a query request to the hospital's material management system through the edge computing gateway, extracting the inner diameter and volume parameters of the infusion tubing to which the current infusion-type life support device belongs based on the consumable barcode. During the installation of consumables, nursing staff input the consumable barcode via a nursing terminal or bedside barcode scanner, and establish a binding relationship between the consumable barcode and the patient and equipment identifiers. The node verification module 20 then queries the corresponding dead space volume parameters of the infusion tubing based on this information. The node verification module 20 also combines this with the set flow velocity vector at the corresponding instruction transition point. Calculate the drug infusion lag period from the dead space of the physical drug drainage tubing to the patient end. : ; In the formula, This represents the dead space volume of the infusion line, measured in milliliters. This represents the newly set flow rate after parameter tuning of the life support equipment, expressed in milliliters per hour. Through division, the node verification module 20 maps the physical distance of the tubing to the time dimension of the drug infusion hysteresis period. .

[0058] The node validation module 20 converts the drug infusion lag period into a physiological response transition time window. To prevent a dead zone in the system's verification logic due to an excessively long lag period when the set flow rate is extremely low, the node validation module 20 configures a maximum protection threshold for the physiological response transition time window. When the calculated drug infusion lag period exceeds the maximum protection threshold, the system forcibly truncates it to the maximum protection threshold. The maximum protection threshold is set with reference to the conventional half-life in clinical pharmacokinetics, and its value range is limited to 30 to 60 minutes. For respiratory support life support devices, the node validation module 20 determines the corresponding physiological response observation time window based on the feedback parameters, waveform characteristics, or alarm status after parameter changes, and can configure the corresponding maximum protection threshold according to the device type. Finally, the node validation module 20 anchors the corrected physiological response transition time window as an entity attribute to the corresponding actual intervention node and outputs it synchronously with the objective intervention set.

[0059] See attached document Figure 5 , Figure 5 This is a flowchart of semantic parsing and subjective statement extraction for handover documents according to an embodiment of the present invention. To transform the unstructured document content entered by medical staff into a data structure that the system can process, the semantic parsing and subjective statement extraction steps for handover documents in this embodiment include the following sub-steps.

[0060] S301, the declaration parsing module 30 intercepts the handover electronic form data from the clinical information system 14 at the end of the handover observation time window. As a preferred method, the declaration parsing module 30 listens for form submissions in the clinical information system 14 via a message queue. When medical staff trigger a submission operation on the nursing terminal interface, the declaration parsing module 30 captures the trigger signal and reads the raw data message generated in the system background. The data message is divided into a structured data area and an unstructured text area. The structured data area contains drop-down selection options and numerical input boxes, while the unstructured text area contains freely written handover notes and a summary of the patient's condition.

[0061] S302, after acquiring the data message, the declaration parsing module 30 parses the structured checkboxes and text fields in the shift handover electronic form, extracts nursing operations and medical order execution events, and assigns semantic timestamps. For the structured data area, the system extracts the corresponding options based on the preset primary key mapping relationship of the data table. For the unstructured text area, the declaration parsing module 30 calls the built-in clinical standard dictionary and regular expression matching engine, where the clinical standard dictionary consists of an operation behavior lexicon and a medical entity lexicon. By recognizing the combined phrases of behavior and entity in the text, the declaration parsing module 30 extracts clinical events such as adjusting the pump rate or pausing the infusion.

[0062] Based on the extracted clinical events, the system assigns them quantified time attributes. The declaration parsing module 30 extracts time strings from the text description using regular expressions. When the time string is identified as representing an absolute time, the system directly converts it into a standard timestamp. If the time string presents a relative time description, the declaration parsing module 30 derives and calculates the semantic timestamp using the following formula. : ; In the formula, Indicates the system submission time of the form. This represents the relative offset duration parsed from the text. To prevent calculation dead zones caused by missing time records in the text, if the time string is not extracted, the parsing module 30 declares that it will execute a fallback strategy, assigning the form creation time as a substitute value to the semantic timestamp. .

[0063] The declaration parsing module 30 encapsulates the extracted clinical events and their corresponding semantic timestamps into declaration clinical nodes, and aggregates nodes within the same observation time window to generate a set of subjective declarations. Preferably, the declaration clinical node includes at least one or more of the following: patient identifier, device identifier or device type, event type, parameter item, parameter value or direction of change, semantic timestamp, and text source location. For word segmentation algorithms and named entity recognition mechanisms in natural language processing, those skilled in the art can implement them using standard text analysis programs. The underlying trie matching and regular expression engine calls are well-known technologies in the field and will not be elaborated upon here.

[0064] S303, another verification dimension of shift handover confirmation lies in obtaining the final state of the equipment at the time of handover. The declaration parsing module 30 extracts the final operating parameters of the equipment from the electronic shift handover form. Specifically, the declaration parsing module 30 locates the equipment parameter verification module at the end of the form and reads the manually confirmed equipment setting parameters one by one, such as the flow rate setting of the micro-infusion pump and the pressure value supported by the ventilator. The declaration parsing module 30 fills the above values ​​into a one-dimensional array in sequence according to the sorting rules of the equipment identification code, generating a subjective declaration final state vector. The subjective declaration final state vector represents the equipment operating state under the subjective perception of the shift handover personnel, providing a comparison benchmark for subsequent differential comparison with the underlying hardware resting state.

[0065] See attached document Figure 6 , Figure 6 This is a logic diagram for cross-dimensional asymmetric differential comparison and early warning distribution according to an embodiment of the present invention. After obtaining the objective intervention set and the subjective statement set, the system verifies the consistency between clinical records and underlying device operation data. In this embodiment, the cross-dimensional asymmetric differential comparison and early warning distribution steps include the following sub-step flow process.

[0066] S401, the differential comparison and early warning module 40 performs cross-time sequence verification. The system confirms whether the physical operations occurring at the device's underlying level are reflected in the nurse's written documentation. Specifically, the differential comparison and early warning module 40 extracts the actual intervention nodes from the objective intervention set and obtains the corresponding objective intervention occurrence time. The attribute matching includes at least patient identifier matching, device identifier or device type matching, event type matching, and parameter item matching. When multiple candidate clinical declaration nodes exist, the differential comparison and early warning module 40 selects the clinical declaration node with the smallest time distance from the actual intervention node as the matching result. When no clinical declaration node meets the matching conditions, it is determined that the actual intervention node is not covered by the subjective declaration. Simultaneously, the differential comparison and early warning module 40 traverses the clinical declaration nodes with matching attributes in the set of subjective declarations and reads the semantic timestamps corresponding to the clinical declaration nodes. To quantify the time difference between the two, the differential comparison and early warning module 40 calculates the time distance function. : ; In the formula, This indicates the time when the objective intervention occurred, as recorded by the underlying device. This indicates the subjective record time extracted from the handover document.

[0067] In obtaining the time distance function Based on the numerical values, the differential comparison and early warning module 40 determines whether the value exceeds a preset time tolerance threshold. The time tolerance threshold is set according to the allowed supplementary recording time guidelines for clinical nursing documents, and its range is limited to 15 to 30 minutes. If... If the time tolerance threshold is exceeded, or if no matching clinical node for the actual intervention node is found in the set of subjective statements, the differential comparison and early warning module 40 determines that there is a risk of missing documents.

[0068] As a preferred approach, the differential comparison and early warning module 40 triggers a terminal blocking mechanism. The system locks the submit button of the handover form on the front-end interface of the nursing workstation, forcibly blocking the handover process. To assist medical staff in tracing the situation on-site, the system displays the underlying waveform of the equipment during the missed period via a pop-up window, prompting nurses to supplement explanations or correct document times. Locking the handover form submit button handles situations of document omissions or inconsistencies in final state; freezing the signature permission for confirming the handover vital signs baseline handles risky transition situations where the physiological response transition time window has not yet ended. These two mechanisms can be triggered independently or in combination. For the pop-up rendering and control locking mechanism of the nursing workstation front-end interface, those skilled in the art can implement it using a standard graphical user interface development framework. Interface event interception and redrawing are well-known technologies in this field and will not be elaborated upon here.

[0069] S402, the differential comparison and early warning module 40 performs a consistency check on the terminal device status. The handover status of intensive care equipment directly affects the continuity and safety of life support treatment for the next shift; verifying the final handover status is a crucial step in preventing medical errors. The differential comparison and early warning module 40 reads the end time of the handover observation window. The discrete-time series is used to extract the values ​​of the resampling command vectors of each device in the discrete-time series, and then sort them according to the device identifier code to generate an objective final state vector. The differential comparison and early warning module 40 calls the subjective declaration final state vector generated in the preceding steps. .

[0070] To prevent computational dead zones caused by inconsistent array dimensions, the difference comparison and early warning module 40 compares the objective final state vector based on the device identification code. With subjective declaration final state vector Perform index alignment. If the device identification codes of the two are inconsistent, the differential comparison and early warning module 40 fills the corresponding position of the missing side with a preset missing status mark to align the dimensions. After alignment, the differential comparison and early warning module 40 performs state deviation calculation to solve for the deviation vector. : ; In the formula, Characterizes the actual physical settings of the underlying hardware at the time of shift handover. The differential comparison and early warning module 40 represents the operating parameters confirmed by the shift handover personnel in the form. For positions marked with missing status indicators, the module does not perform numerical difference calculations but directly generates a record of missing equipment status conflicts. For positions where both values ​​are valid, the module then performs status deviation calculations. The module determines the deviation vector item by item. The system checks whether each component exceeds the allowable error of the equipment. The allowable error is configured independently based on the hardware range and calibration accuracy of different medical instruments, and its value range is set to 2% to 5% of the instrument's full scale. When a component exceeds the limit, the differential comparison and early warning module 40 triggers a misalignment alarm. The differential comparison and early warning module 40 packages the misalignment alarm along with the deviation vector and distributes it to the mobile terminal and ward management dashboard of the relieving nurse through the hospital's local area network, prompting the relieving staff to verify the equipment parameters at the bedside.

[0071] See attached document Figure 7 , Figure 7This is a flowchart of a closed-loop traceability and cross-shift dynamic risk control lock process for consumables according to an embodiment of the present invention. After completing the comparison and verification of the document declaration and the underlying equipment data, the system enters the physical verification and medication safety management stage for clinical supplies. In this embodiment, the closed-loop traceability and cross-shift dynamic risk control lock process for consumables includes the following sub-steps.

[0072] S501, the traceability and risk control module 50 performs closed-loop accounting of the remaining physical quantity of infusion-type life support medications. To accurately determine the true quantity of available supplies at the time of shift handover, the traceability and risk control module 50 retrieves the medical order execution form of the currently handing-over patient through the hospital information system port, extracting the initial nominal volume of the medication pack for the infusion tubing corresponding to the infusion-type life support equipment. .

[0073] In one implementation, the traceability risk control module 50 also reads the cumulative actual infusion volume of the drug solution package since the moment it was started. ;

[0074] In another implementation, the system has already recorded the initial remaining amount of the medicine pack at the start of the current shift. At that time, the risk control module accumulates the actual physical work feedback points within the current shift's observation time window to obtain the total cumulative consumption for the shift. Based on this, the traceability risk control module 50 calculates the physical balance of the medicine at the shift handover time using the following formula. .

[0075] When the cumulative actual infusion volume since the drug package was put into use is available: ; When the initial surplus of the class is used as the benchmark: ; In the formula, This indicates the initial total volume of the medication prescribed by the doctor. This indicates the cumulative actual outflow volume of the medication package since its commissioning. This indicates the remaining amount of medicine recorded at the start time of the current shift. This represents the cumulative actual outflow volume within the current shift. By subtracting volumes, the risk control module 50 obtains objective and unaffected material surplus data.

[0076] S502, based on clearly defined physical surplus indicators, to prevent the risk of empty medications due to the ward rounds being missed by incoming staff, the traceability risk control module 50 performs logical analysis of cross-shift dynamic risk control locks. For infusion-type life support equipment, the traceability risk control module 50 extracts the set flow rate of the corresponding equipment at the end of the shift handover. And calculate the estimated depletion time in conjunction with the physical surplus. : ; In the formula, This represents the physical surplus obtained from the previous step. This indicates the current set pumping speed of the equipment. This is to prevent issues when the injection pump is in a shutdown maintenance state. The division-by-zero dead zone caused by an integer equal to zero is addressed by tracing the boundary bypass logic configured in the risk control module 50. If... The estimated time for the risk control module to run out is zero. The value is directly assigned to the system's preset infinite safety constant.

[0077] Combined with the calculated estimated exhaustion time The risk control module 50 compares the result with the set safe connection threshold. The safe connection threshold is set based on the standard nursing rounds interval in the intensive care unit, with a range of 1 to 2 hours.

[0078] When the expected exhaustion time When the flow rate falls below the safe continuity threshold, the traceability risk control module 50 triggers a cross-shift dynamic risk control lock. As a preferred method, the physical implementation of the cross-shift dynamic risk control lock includes two interception commands. When the corresponding infusion-type life support equipment has an authorized remote control interface, the traceability risk control module 50 issues a permission contraction command to the corresponding infusion-type life support equipment through the edge computing gateway, locking the operation permission to increase the flow rate on the equipment control panel to prevent accidental operation at the beginning of the shift from accelerating the depletion of the medicine.

[0079] When the corresponding infusion-type life support equipment only has a data reading interface and not a remote control interface, the traceability risk control module 50 implements software-side permission prompts, process blocking, or high-risk confirmation mechanisms at the human-machine interface of the nursing workstation front end, mobile terminal, or edge computing gateway, instead of directly controlling the equipment itself. Simultaneously, the traceability risk control module 50 displays a mandatory intervention interface at the nursing workstation front end, requiring the outgoing and incoming nurses to use their work badges to perform double-person scanning confirmation at the bedside. Only after confirming the requisition of the spare medication can the terminal be unlocked.

[0080] S503, the traceability risk control module 50 generates a global closed-loop traceability log. Specifically, the traceability risk control module 50 structurally assembles the actual intervention nodes, subjective declaration sets, deviation vectors, and risk control lock trigger records generated in the preceding steps, and adds an absolute timestamp of the shift handover confirmation time. The traceability risk control module 50 uses a hash algorithm to calculate the digest feature value of the structured data packet, generates a tamper-proof closed-loop traceability log, and uploads it to the hospital's core database. The closed-loop traceability log provides an objective chain of evidence supported by underlying equipment data for subsequent medical dispute evidence collection and nursing quality control. For the underlying algorithms of log encryption and hash value generation, those skilled in the art can use standard cryptographic function libraries for implementation; the data encryption and storage mechanism is a well-known technology in the field and will not be elaborated upon here.

[0081] The following example demonstrates the closed-loop processing flow of the system, using a real shift handover scenario in the intensive care unit.

[0082] The observation scenario was set during shift handover for patients with severe pneumonia complicated by septic shock. A multi-parameter monitor and a microinfusion pump were placed at the patient's bedside. The microinfusion pump continuously infused vasoactive drugs to maintain target blood pressure. Within a certain timeframe before the end of the shift handover observation window, such as 15 minutes before the handover confirmation point, if the handover nurse observed a decreasing trend in the patient's mean arterial pressure, the flow rate of the medication was increased from the original 4 ml / h to 8 ml / h on the microinfusion pump's physical control panel.

[0083] During this operation, the synchronous sampling module acquires changes in the underlying data frame, and the node verification module calls the device capability profile matrix to perform first-order differential detection to locate the flow rate command abrupt change point. The system combines the actual physical work feedback integral within the dynamic verification integration window to verify the underlying hardware pumping action and generate an actual intervention node with an absolute timestamp. The system extracts the pre-entered dead space volume parameters of the current infusion line and calculates, based on the adjusted flow rate vector, that the drug infusion hysteresis period for this rate adjustment command is 22 minutes. Therefore, the system anchors the endpoint of the physiological response transition time window 22 minutes after the intervention node occurs.

[0084] At the end of the shift handover observation time window, the handover nurse submits an electronic handover form at the front end of the nursing workstation. The declaration parsing module intercepts the message and extracts the text description and parameter status. If the nurse completely misses the aforementioned speed adjustment operation record due to subjective memory bias, or mistakenly records the time as 40 minutes before the handover, which is significantly different from the actual time, the differential comparison and early warning module will identify that the time deviation exceeds the system's set time tolerance threshold when performing a time-series cross-check of the objective intervention set and the subjective declaration set, and then trigger the terminal blocking mechanism. The system intercepts the form flow on the front end interface, forcibly pops up the flow waveform and blood pressure fluctuation slices of the micro-infusion pump before and after the actual intervention time, and prompts the nurse to correct the document time.

[0085] Furthermore, the calculated infusion lag period was 22 minutes, which is greater than the remaining 15 minutes between the rate adjustment operation and the shift handover. This caused the actual handover confirmation time to fall directly within the physiological response transition time window. This indicates that the additional medication delivered by increasing the pump rate has not yet fully crossed the dead space of the tubing to reach the patient's vein, and the systemic vital signs are in a non-steady state. Based on this, the risk control module determined that there was a risk blind spot across shifts, automatically generated a cross-shift intervention connection identifier in the form, froze the confirmation signature permission for the current shift's baseline vital signs, and alerted the incoming shift staff that the patient was in a period of pharmacokinetic fluctuation. The signature restriction was only lifted after the system determined that the transition time window had completely expired and the blood pressure baseline had stabilized again. Simultaneously, the traceability risk control module calculates the physical remaining amount of the corresponding medication package. If the calculated remaining medication is expected to be exhausted within 45 minutes at a flow rate of 8 ml per hour, and this value is lower than the safe continuation threshold of 1 hour, the system triggers a cross-shift dynamic risk control lock. The system locks the operation permission of the micro-infusion pump to increase the flow rate through the edge computing gateway, and requires the two people in the handover to use smart work badges to scan the code at the bedside to confirm the medication preparation.

[0086] To verify the effectiveness of this system in a clinical setting, a field comparative trial was conducted in the intensive care unit of a tertiary hospital. Subjects were randomly divided into a control group and an experimental group based on physical isolation within the ward. The control group used a conventional electronic document-based manual shift handover method, while the experimental group ran the intelligent verification, error warning, and traceability management system provided by this invention. Evaluation indicators included average patient handover time, omission rate of key clinical event records, inconsistency rate in equipment handover final state, and frequency of high-risk consumable warnings.

[0087] The anonymized statistical data was obtained from the hospital information system backend, and the summary and comparison results are shown in the table below:

[0088] Statistical table of clinical comparative experiments on the effectiveness of traditional shift handover models and the application of this system:

[0089] Experimental data shows that the system, through unified sampling of underlying equipment data and verification of objective entity nodes, reduces manual transcription and secondary manual verification, shortening the time required for shift handover. The cross-dimensional asymmetric differential comparison mechanism effectively intercepts document errors caused by subjective memory bias, reducing the rate of missed critical events. State consistency verification and dynamic risk control locks eliminate the risk of equipment final state misalignment, significantly reducing the frequency of high-risk medical events caused by empty consumable supplies, and verifying the engineering feasibility of the system in ensuring the safety of continuous life support treatment in the ICU.

[0090] See attached document Figure 8 The figure shows the mapping distribution of discrete feature variables extracted from multi-parameter monitors and life support equipment by the system's underlying processing framework, demonstrating the topological constraint association process between equipment operation commands and underlying physical feedback at different time granularities.

[0091] See attached document Figure 9 The figure uses multidimensional scaling transformation to map the clinical monitoring data of the control group and the experimental group to a low-dimensional state space, presenting the convergence state and blocking boundary characteristics of the system for abnormal sample nodes such as document recording deviation, equipment final state misalignment and potential drug discontinuation risk, intuitively reflecting the robustness of the system algorithm model at the risk interception level.

[0092] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent verification, error early warning, and traceability management of ICU nursing shift handover information, characterized in that, Includes the following steps: S10: Obtain the vital sign response vector output by the monitor, obtain the digital command vector and physical feedback vector output by the life support device, perform unified frequency resampling based on the global time base, and generate a time-axis aligned discrete time series. S20, perform differential detection to locate command jump points on the digital command vector in the discrete time series, construct a dynamic confirmation integral window to calculate the actual physical work feedback integral, generate actual intervention nodes to form an objective intervention set when the actual physical work feedback integral meets the error condition, calculate the physiological response transition time window according to the equipment type and anchor it to the actual intervention node. S30, extract text record events and timestamps from the shift handover form to generate a set of subjective statements, extract equipment operating parameters from the shift handover form to generate a final state vector of subjective statements, and use the parameters corresponding to the discrete time series at the shift handover time to generate an objective final state vector. S40, calculate the time deviation distance between the associated nodes in the objective intervention set and the subjective declaration set, and trigger a terminal blocking warning when the time deviation distance exceeds the time deviation tolerance threshold; By comparing the objective final state vector with the subjective declared final state vector, a final state conflict warning is triggered when there is a parameter difference between the objective final state vector and the subjective declared final state vector. S50: When the handover time falls within the physiological response transition time window corresponding to the actual intervention node, freeze the confirmation signature permission of the handover form until the physiological response transition time window is exhausted. In step S20, differential detection is performed on the digital command vector in the discrete time series to locate the command jump point, the actual physical work feedback integral is calculated, and an actual intervention node is generated when the actual physical work feedback integral meets a preset error condition, including: Calculate the instruction difference between digital instruction vectors at adjacent discrete time nodes. When the instruction difference is greater than the minimum response step size of the device and the duration of the non-zero state of the digital instruction vector meets the set duration condition, the starting time of the state change of the digital instruction vector is located as the instruction jump point. The theoretical total work done is obtained by multiplying the accumulated digital command vector values ​​within the dynamic verification integration window by the reference sampling period. Extract the increment of the physical feedback vector within the dynamic verification integration window as the actual physical work feedback integral; Determine whether the ratio of the absolute value of the difference between the theoretical total work and the actual physical work feedback integral to the theoretical total work falls within the tolerance error zone condition. If the tolerance error zone condition is met, then generate an actual intervention node. In step S20, the physiological response transition time window is calculated based on the device type, including: When the life support device is an infusion-type life support device, the dead cavity volume parameter of the infusion pipeline to which the infusion-type life support device belongs is extracted, and the dead cavity volume parameter is divided by the set flow rate value corresponding to the command jump point to calculate the drug infusion hysteresis period. Configure a maximum protection threshold. When the drug infusion lag period exceeds the maximum protection threshold, the drug infusion lag period is truncated to the maximum protection threshold. The truncated drug infusion lag period is used as the physiological response transition time window. The initial nominal volume of the drug solution pack in the corresponding infusion line when the life support device is an infusion-type life support device; Following S50, a consumables closed-loop traceability step is also included: The initial nominal volume of the drug solution pack in the corresponding infusion line when the life support device is an infusion-type life support device; The total cumulative consumption is calculated by summing the actual physical work done during the shift handover period; The physical balance of the medicine at the handover time is obtained by subtracting the total cumulative consumption from the initial nominal volume. The actual intervention nodes, subjective declaration sets, and state missing conflict records are structurally assembled and timestamped. A hash algorithm is used to calculate the summary feature value to generate a closed-loop traceability log. Following the consumables closed-loop traceability step, the process also includes triggering a cross-shift dynamic risk control lock: Divide the physical surplus by the set flow rate at the shift handover time of the infusion-type life support equipment to calculate the estimated depletion time; When the expected exhaustion time is less than the safe continuity time threshold, a permission contraction command is issued to the infusion life support device with a remote control interface, and the operation permission to increase the set flow rate is locked on the control panel of the infusion life support device.

2. The method for intelligent verification, error early warning, and traceability management of ICU nursing handover information according to claim 1, characterized in that, In step S10, a unified frequency resampling is performed on the symptom response vector, digital command vector, and physical feedback vector based on a global time reference to generate a time-axis aligned discrete-time series, including: Receive data frames containing vital sign response vectors, digital command vectors, and physical feedback vectors; strip the original timestamps from the data frames; and append global arrival timestamps based on the global time base. A discrete time node sequence is generated based on the reference sampling period. Linear interpolation resampling is performed on the vital sign response vector, and zero-order hold resampling is performed on the digital command vector and physical feedback vector.

3. The method for intelligent verification, error early warning, and traceability management of ICU nursing handover information according to claim 1, characterized in that, In step S30, text record events and timestamps are extracted from the electronic shift handover form to generate a set of subjective statements, including: The built-in clinical standard dictionary is invoked to identify text record events in the unstructured text of the shift handover electronic form; Extract time strings from unstructured text, convert absolute time strings to standard timestamps, and calculate timestamps for relative time strings by subtracting the relative offset duration from the submission time of the shift handover electronic form.

4. The method for intelligent verification, error early warning, and traceability management of ICU nursing handover information according to claim 1, characterized in that, In step S40, calculating the time deviation distance between associated nodes in the objective intervention set and the subjective statement set includes: Traverse the set of subjective statements and match clinical nodes with the same device identifier and event type as the actual intervention node. When there are multiple clinical nodes with statements, select the clinical node with the smallest time interval between the occurrence time and the actual intervention node as the matching result. The time deviation distance is calculated by subtracting the timestamp of the matching result from the actual time of occurrence of the intervention node.

5. The method for intelligent verification, error early warning, and traceability management of ICU nursing handover information according to claim 1, characterized in that, In step S40, comparing the objective final state vector with the subjective declared final state vector includes: Perform index alignment between the objective final state vector and the subjective declared final state vector based on the device identification code; For locations where device identification codes are mismatched, fill in missing status markers to generate status missing conflict records; Numerical difference calculation is performed on the alignment position of the device identification code in the objective final state vector and the subjective declared final state vector. When the numerical difference exceeds the device calibration accuracy range, a final state conflict warning is triggered.

6. The method for intelligent verification, error early warning, and traceability management of ICU nursing handover information according to claim 1, characterized in that, In step S40, triggering a terminal blocking warning when the time deviation distance exceeds a preset timing deviation tolerance threshold includes: The handover electronic form submission process is locked on the front-end interface of the nursing workstation, and the time-series slice data of vital sign response vectors before and after the actual intervention point are retrieved and displayed.

Images