A method and system for assessing the risk of catheter tip displacement

Abstract

This invention discloses a method and system for assessing catheter tip displacement risk. It collects patient positional change frequency data and historical catheter positioning deviation records. Risk factor sets are extracted through rhythm disorder feature analysis. Combining the risk factor sets with historical deviation records, displacement risk characteristic factors are extracted and risk contribution assessments are performed to generate a risk intervention plan. Scoring indicators are extracted from the risk intervention plan to construct a displacement scoring rule library. A set of sensitive displacement parameters is identified and archived to generate a displacement risk profile. Scoring item association analysis is performed on the displacement risk profile, and a graded displacement assessment sequence is formed through drift trend extraction and weighted labeling. Displacement probability calculation is performed on the graded displacement assessment sequence. Catheter tip displacement risk assessment results are generated through stratified monitoring of fluctuation amplitude and matching of re-examination frequency, achieving dynamic assessment of catheter tip displacement risk throughout the entire process.

Description

Technical Field

[0001] This invention relates to the field of clinical nursing decision support technology, and in particular to a method and system for assessing the risk of catheter tip displacement. Background Technology

[0002] Indwelling catheters are an important clinical treatment for critically ill, postoperative, and long-term bedridden patients. The accuracy of catheter tip positioning directly affects drug infusion efficacy and treatment safety. During hospitalization, the risk of catheter tip displacement persists due to changes in consciousness, passive repositioning, or nursing interventions. However, current nursing assessment systems mainly rely on periodic imaging reviews and subjective judgment, lacking a systematic integration of individual patient behavioral characteristics and historical catheterization records, making it difficult to identify high-risk conditions before displacement occurs.

[0003] Significant individual differences exist among patients in terms of disease severity, nursing compliance, and catheter type. A standardized assessment process cannot accurately differentiate the actual risk level of each patient, leading to a persistent mismatch between monitoring resource allocation and the true risk distribution. When a patient's risk status dynamically evolves over time, the assessment results lack a quantitative description of the drift trend, making it difficult for nursing staff to determine whether the current risk is on an accelerating deterioration trajectory. Adjustments to the frequency of follow-up examinations also lack objective support, ultimately affecting the timeliness and targeted nature of catheter safety management. Summary of the Invention

[0004] This invention discloses a method and system for assessing catheter tip displacement risk. By integrating patient postural behavior characteristics and historical catheter displacement data, a complete assessment closed loop is established, from risk factor extraction, intervention plan generation, risk file archiving to drift trend assessment and stratified early warning, thereby realizing dynamic quantification and individualized monitoring resource allocation for catheter tip displacement risk.

[0005] The first aspect of this invention provides a method for assessing the risk of catheter tip displacement, comprising the following steps:

[0006] Collect patient position change frequency data and catheter historical positioning deviation records, and perform clinical risk factor extraction to form a risk factor set based on the rhythm disorder characteristics of the position change frequency data;

[0007] Using the risk factor set and the catheter historical positioning offset record, displacement risk feature factors are extracted. The critical displacement risk contribution of the displacement risk feature factors is evaluated to generate a risk contribution level. Based on the risk contribution level, factor weight difference mapping is performed to generate a risk intervention plan.

[0008] The risk intervention plan is subjected to scoring indicators to form a shift scoring rule library. The shift scoring rule library is used to identify a set of sensitive shift parameters. Based on the set of sensitive shift parameters, a benchmark archiving and storage is performed to generate a shift risk profile.

[0009] The displacement risk profile is analyzed by performing a scoring item association parsing to extract the displacement risk association chain. Based on the displacement risk association chain, the temporal drift features are extracted to generate a trend-weighted identifier. The trend-weighted identifiers are then arranged according to the risk priority to form a graded displacement assessment sequence.

[0010] The displacement probability is calculated to obtain the displacement risk probability value of the graded displacement assessment sequence. The stratified monitoring range is defined based on the fluctuation range of the displacement risk probability value to generate a stratified early warning identifier. The re-examination frequency is matched based on the stratified early warning identifier and the risk factor set to generate the catheter tip displacement risk assessment result.

[0011] A second aspect of this invention provides a catheter tip displacement risk assessment system, comprising:

[0012] The risk factor module is used to collect patient position change frequency data and catheter historical positioning deviation records, and to perform clinical risk factor extraction to form a risk factor set based on the rhythm disorder characteristics of the position change frequency data;

[0013] The risk assessment module is used to extract displacement risk characteristic factors using the risk factor set and the catheter historical positioning offset record, perform critical displacement risk contribution assessment on the displacement risk characteristic factors to generate risk contribution level, and perform factor weight difference mapping based on the risk contribution level to generate risk intervention plan;

[0014] The file storage module is used to extract scoring indicators from the risk intervention plan to form a shift scoring rule library, use the shift scoring rule library to identify a set of sensitive shift parameters, and perform benchmark archiving and storage based on the set of sensitive shift parameters to generate a shift risk file;

[0015] The trend arrangement module is used to perform scoring item association parsing on the displacement risk file to extract displacement risk association chains, extract time-series drift features based on the displacement risk association chains to generate trend weighted identifiers, and arrange the trend weighted identifiers according to the risk priority to form a graded displacement assessment sequence.

[0016] The stratified early warning module is used to perform displacement probability calculation on the stratified displacement assessment sequence to obtain displacement risk probability value, delineate the stratified monitoring range based on the fluctuation range of the displacement risk probability value to generate a stratified early warning identifier, and perform re-examination frequency matching based on the stratified early warning identifier and the risk factor set to generate catheter tip displacement risk assessment result.

[0017] The beneficial effects of this invention are reflected in the following points: 1. It uses the rhythmic disorder segment of the patient's postural change frequency data as an indirect objective indicator of the degree of consciousness impairment, avoiding the limitations of relying on subjective assessment scales; it introduces the catheter duration decay resistance feature in the risk contribution assessment stage, distinguishing between two types of factors: the continuous accumulation of risk with catheter duration and natural decay, solving the problem of long-term high-risk factors being diluted due to traditional equal-weighted superposition; it verifies and screens the intervention configurations that can be reused through retrospective verification of the incidence of displacement events, so that the intervention plan is based on empirical evidence tested by historical effectiveness rather than simply relying on theoretical inference. 2. It uses the patient's individual admission baseline instead of the mean of the same group as the deviation comparison benchmark, fundamentally eliminating misjudgments caused by systematic deviation of the group baseline due to individual physical differences. This is the core design that distinguishes it from existing scoring tools; after classifying the difference parameters, it traces their co-occurrence patterns before historical deviation events, and compares the current parameter deviation pattern with the historical empirical path, upgrading risk prediction from statistical probability inference to individualized historical pattern matching. The traceability path in the file ensures that each risk judgment conclusion has verifiable basis for its formation. 3. By analyzing the causal chain of scoring items, the leading node of risk transmission can be clearly identified, which can obtain a longer early warning lead time compared to monitoring only the final state indicators; the combination of sudden change segment identification and segmented slope breakpoint location separates instantaneous jumps from trend-based accelerated deterioration, solving the deficiency of traditional overall regression in masking local acceleration stages; the shift probability formula simultaneously incorporates two inputs: score drift trend and risk factor confidence, and correlates probability fluctuations with nursing operation types to achieve patient stratification based on operation disturbance patterns, making the allocation of monitoring resources highly matched with the spatiotemporal distribution of real risks. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a method for assessing the risk of catheter tip displacement according to the present invention.

[0019] Figure 2 This is a structural block diagram of a catheter tip displacement risk assessment system according to the present invention. Detailed Implementation

[0020] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0021] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0022] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0023] The technical solutions of the embodiments of this application will be described below.

[0024] like Figure 1 As shown, this embodiment of the invention provides a method for assessing the risk of catheter tip displacement, including the following steps S110-S150:

[0025] Step S110: Collect patient position change frequency data and catheter historical positioning deviation records, and perform clinical risk factor extraction based on the rhythm disorder characteristics of the position change frequency data to form a risk factor set.

[0026] Specifically, patient position change frequency data and catheter historical positioning offset records are collected. Position change frequency data is continuously collected via a bedside pressure sensor, which triggers recording when the patient's weight distribution changes. Each trigger extracts the time of occurrence of the event, which serves as the temporal basis for event counting and interval calculation within the sliding window in subsequent rhythm pattern analysis. Catheter historical positioning offset records are extracted from the hospital information system, covering the offset direction, offset distance, and timestamp of each imaging follow-up. Offset direction is categorized into head-end offset and foot-end offset. Both position change frequency data and catheter historical positioning offset records are affixed with a patient identifier and a collection timestamp. The two data streams are linked via the patient identifier to achieve patient-level data binding. Position change frequency data from different collection periods for the same patient are spliced ​​into a continuous sequence in chronological order. When the interval between spliced ​​periods exceeds 6 hours, a data interruption marker is added. This interruption marker triggers breakpoint processing during the rhythm pattern analysis phase to avoid misjudgments in rhythm calculations across intervals. In the history of catheter positioning deviation, adjacent records with a follow-up interval of more than 72 hours are considered as long-interval follow-up pairs. For example, if a patient did not have an imaging follow-up for three consecutive days over the weekend, the catheter tip deviation distance was found to be significantly larger on Monday's follow-up than on Thursday's last follow-up. However, it is impossible to determine at what time the deviation occurred or at what rate. Such records are marked as low temporal resolution records.

[0027] In some embodiments, the step of extracting clinical risk factors based on the rhythm disorder characteristics of the body position change frequency data to form a risk factor set includes: performing rhythm pattern analysis on the body position change frequency data to obtain rhythm feature sequences; identifying disordered rhythm segments in a state of impaired consciousness based on the rhythm feature sequences to generate disordered rhythm identifiers; performing clinical risk factor association matching based on the disordered rhythm identifiers to form a factor association table; and performing standardized integration based on the factor association table to form a risk factor set.

[0028] Rhythmic pattern analysis was performed on the frequency data of body position changes to obtain rhythmic characteristic sequences. The rhythmic pattern analysis used the time series of body position change frequency data as input, and employed a sliding window approach to count the frequency of body position change events per unit time. The window length was set to 1 hour, with a step size of 15 minutes. The temporal arrangement of the number of change events within the window reflected the rhythmic pattern of body position changes. Taking ordinary hospitalized patients as an example, during the daytime awake period, bedside pressure sensors typically recorded 8 to 15 body position change triggers per hour. After falling asleep at night, the frequency decreased to 2 to 5 times per hour and showed a regular low-frequency distribution. The mean frequency and maximum interval duration exhibited periodic fluctuations at the day-night transition point. However, in patients under continuous postoperative sedation, body position change events almost completely disappeared within several consecutive windows. The mean frequency approached zero, while the maximum interval duration continuously expanded to several hours, and the frequency variance simultaneously approached zero. The combined pattern of these three characteristics showed a significantly different and distinguishable difference from the normal low-frequency state at night. The rhythmic feature sequence arranges the feature quantities of each sliding window in chronological order. The frequency mean reflects the overall activity level of postural changes within that window, the frequency variance reflects the stability of this activity, and the maximum interval captures long periods of stillness without postural changes. The frequency mean, frequency variance, and maximum interval jointly describe the rhythmic pattern type within the window. Windows with data interruption annotations in the postural change frequency data have interruption inheritance annotations added to the corresponding positions in the rhythmic feature sequence. The three feature quantities of the interruption inheritance-annotated windows are still included in the calculation, but only the valid data segments before and after the interruption are used for separate statistics; the average is not merged across interruption points.

[0029] Disordered rhythmic segments in states of altered consciousness are identified and disordered rhythm markers are generated based on rhythmic feature sequences. The core characteristic that distinguishes postural change rhythms in states of altered consciousness from normal low-frequency sleep is the abnormal synchronization of the duration and frequency variance of the resting segment. In normal low-frequency states, the frequency variance remains relatively stable, while in cases of passive postural insufficiency due to altered consciousness, the frequency variance approaches zero and the maximum interval duration is consistently larger. When the mean frequency in the rhythmic feature sequence is lower than the normal baseline mean by more than 50% for three consecutive windows, and the maximum interval duration synchronously exceeds twice the normal baseline, the corresponding window segment is identified as a candidate segment for disordered rhythm. The normal baseline is determined by statistical values ​​of postural change frequency data within 72 hours prior to admission for the same patient; if pre-admission data is insufficient, the mean of the same age group in the same ward is used as a substitute. Disordered rhythm identifiers define the time range of each candidate disordered rhythm segment using start and end window numbers. The disorder depth of the segment is quantified by the mean frequency and mean maximum interval duration within the segment. Segments with lower mean frequency and higher mean maximum interval duration are more disordered. For patients who have undergone prolonged sedation postoperatively, the mean frequency of disordered rhythm identifiers during the sedation period is close to zero, and the mean maximum interval duration is far beyond the normal range. The disorder depth of this segment is at the highest level in disordered rhythm identifiers. When generating disordered rhythm identifiers for segments containing interrupted inheritance annotation windows in rhythmic feature sequences, the threshold for the number of consecutive disorder judgment windows is relaxed from 3 to 2 to avoid missing true disordered segments due to insufficient effective segment length caused by data interruptions.

[0030] A factor association table is generated by matching clinical risk factors based on disordered rhythm identifiers. This matching process matches the descriptive features of each disordered segment within the disordered rhythm identifier against a pre-built risk factor knowledge base. The knowledge base stores catheter displacement-related clinical risk factors categorized by disorder type, covering four main sources: degree of consciousness impairment, lack of postural restraint, resistance to active movement, and use of sedatives. Each source corresponds to several specific risk factor entries. The more severe the disorder in the disordered rhythm identifier, the wider the range of risk factor entries triggered. Mildly disordered segments only trigger factors related to frequent turning leading to changes in catheter tension, while deeply disordered segments simultaneously trigger factors related to loss of postural control and ineffective restraint measures. The factor association table records the matching relationship between each risk factor entry and its corresponding disordered rhythm identifier segment. Each matching record locates the specific risk factor entry with its factor number and associates it with the corresponding disordered rhythm identifier segment with its trigger segment number. The matching confidence level measures the degree of agreement between the two, which is determined by the similarity between the disordered segment description features and the typical trigger features of the factor in the knowledge base. In a certain patient, the mean maximum interval duration in the disordered rhythm identifier during postoperative sedation was consistently large, which highly matched the typical trigger features of the sedation drug use factor in the knowledge base, and the corresponding matching confidence level was close to perfect. The disordered rhythm identifier matching results in the corresponding time period of the low temporal resolution records in the catheter history positioning offset records are marked with low-resolution annotations in the factor association table.

[0031] A risk factor set is formed by standardizing and integrating the factor association table. Standardization and integration involves deduplicating and merging all risk factor entries in the factor association table by factor number. When the same factor entry appears multiple times in the factor association table, the weighted average of the matching confidence scores of each occurrence is taken as the overall confidence score of the factor. The weight is determined by the depth of disorder in the corresponding triggering segment; the deeper the disorder, the higher the weight of the matching record triggered by the segment. Entries in the factor association table with a matching confidence score below 0.4 are marked as low-confidence entries during the standardization and integration phase. Low-confidence entries are not directly included in the main body of the risk factor set but are separately summarized as candidate appendix entries. Candidate appendix entries are upgraded to be included in the main body when similar deviations are found in the catheter history positioning deviation records. The overall confidence score of each factor entry in the risk factor set is used to quantify the clinical significance of the factor in the current patient state. A list of disordered segment numbers from the associated sources is also compiled to support tracing which disordered rhythm segments triggered the factor. The factor category is labeled to indicate which of the following four sources it belongs to: degree of consciousness impairment, lack of postural restraint, resistance to active movement, or use of sedative drugs. The overall confidence level of low-resolution labeled entries in the factor association table is multiplied by a reduction factor of 0.8 during integration. Entries with a confidence level still higher than 0.4 after reduction are normally included in the risk factor set, while entries with a confidence level lower than 0.4 after reduction are downgraded to candidate appendix entries. The risk factor set is finally arranged in descending order of overall confidence level, with the factor entry with the highest confidence level at the beginning of the set.

[0032] Step S120: Extract displacement risk characteristic factors using the risk factor set and catheter historical positioning offset records; conduct a critical offset risk contribution assessment on the displacement risk characteristic factors to generate a risk contribution level; and generate a risk intervention plan based on the factor weight difference mapping according to the risk contribution level.

[0033] Specifically, displacement risk characteristic factors are extracted using a risk factor set and catheter historical positioning offset records. The overall confidence level of each factor entry in the risk factor set and the factor category label are used as search criteria. Historical offset events corresponding to the factor category are located in the catheter historical positioning offset records. A temporal association is established when the occurrence time of the offset event overlaps with the time period of the corresponding factor entry's source disorder segment. For example, the source disorder segment of a patient's consciousness impairment factor covers the period of postoperative sedation. A displacement event occurred during this time period in the catheter historical positioning offset records, and the offset time falls within the disorder segment, establishing a valid association. This indicates that the factor historically did indeed co-occur with catheter offset events, providing empirical support. Factor entries with high overall confidence and temporal association in the risk factor set are extracted as displacement risk characteristic factors. Factor entries supported only by overall confidence but without historical offset event association are marked as purely theoretical risk factors. Purely theoretical factors are included in the displacement risk characteristic factors but with an additional annotation indicating no historical corroboration. In catheter history positioning deviation records, deviation events with deviation distances exceeding the 75th percentile of deviation distances in the same patient group are defined as significant deviation events. Factor entries that have a time association with significant deviation events are marked as significant deviation association factors in the displacement risk characteristic factors. The significant deviation association factors are supplemented with two statistics: the historical significant deviation association count and the maximum historical deviation distance, reflecting their historical co-occurrence density with large-amplitude catheter deviation events. Candidate appendix entries in the risk factor set are upgraded and incorporated into the main body of the displacement risk characteristic factors when similar deviation histories are found in the catheter history positioning deviation records. The overall confidence level of upgraded entries is taken as the original value before reduction to restore their actual reliability after historical confirmation. The displacement risk characteristic factors are arranged in descending order of both overall confidence level and historical association strength.

[0034] In some embodiments, the step of assessing the critical offset risk contribution of the displacement risk characteristic factor to generate a risk contribution level includes: extracting the offset critical record of the displacement risk characteristic factor; identifying the catheter duration attenuation resistance feature based on the offset critical record to obtain a critical proximity value; classifying the critical proximity value to generate a contribution classification result; and determining the risk contribution level based on the contribution classification result.

[0035] The critical offset records for displacement risk characteristic factors are extracted. The extraction of critical offset records involves historical offset events in the catheter historical positioning offset records that are temporally correlated with each displacement risk characteristic factor. Events whose offset distance reaches or exceeds a clinical safety threshold are selected as critical offset events. The clinical safety threshold is defined as the minimum distance the catheter tip must offset to the boundary of the target location. This threshold is pre-set in the system parameter library based on catheter type and placement location. The critical offset records are constructed around each critical offset event. The catheter placement duration reflects the cumulative impact of the catheter's in-vivo time on the critical offset risk. The overall confidence level of the corresponding displacement risk characteristic factor measures the clinical confirmation strength of that factor. The duration of abnormal postural rhythm within 72 hours prior to the offset event reflects the depth of the disorder that triggered the critical offset. These elements collectively support subsequent attenuation resistance feature identification and critical approach value calculation. Factors without historical corroboration in the displacement risk characteristic factors cannot have their critical displacement records extracted from the catheter historical positioning displacement records. For example, a first-time catheter placement patient had neither personal historical displacement data nor any displacement-related records for the active activity resistance factor in their risk factor set prior to this hospitalization. Therefore, the system instead extracted statistical critical records from a similar patient group—patients of similar age, with the same catheter type, and with a history of active activity resistance—as substitutes. These substitute records were further annotated with group statistical labels to distinguish them from individual measured records. In the critical displacement records, each critical displacement event is arranged in ascending order of catheter placement duration at the time of occurrence, with the shortest placement duration event listed first. For instance, a patient experienced a critical displacement event less than 48 hours after catheter placement due to frequent turning, corresponding to a shorter placement duration in the first critical displacement record. This record structure directly reflects the critical trigger distribution characteristics of this factor in the early stages of catheter placement.

[0036] Based on the critical offset records, the critical approach value was obtained by identifying the catheter duration decay resistance feature. The catheter duration decay resistance feature describes the ability of a certain displacement risk factor to continuously drive the catheter towards the critical offset state as the catheter duration increases. Factors with strong decay resistance maintain a high risk contribution even with longer catheter durations, while the risk contribution of factors with weak decay resistance decreases rapidly with increasing catheter duration. The catheter duration and corresponding offset distance of each critical offset event in the critical offset records form a scatter plot. Linear regression was performed on the scatter plot, and the regression slope γ reflects the trend of the offset distance changing with the catheter duration. A positive γ indicates that the offset distance tends to increase with increasing catheter duration, while a negative γ indicates that the offset risk is alleviated with increasing catheter duration. The larger the absolute value of the slope, the more significant the trend. The critical approach value T is determined by the formula T = γ_norm × C_conf × W_abnormal, where γ_norm is the normalized value obtained by dividing the regression slope by the maximum absolute value of the slope in the same patient group, ranging from -1 to 1; C_conf is the comprehensive confidence level of the displacement risk characteristic factor; and W_abnormal is the normalized weight of the mean duration of abnormal postural rhythm in the deviation critical record. The product of these three terms reflects the quantification degree of the factor driving the catheter to approach the critical deviation under actual catheter placement. The critical approach value corresponding to the deviation critical record labeled by the population statistics is multiplied by a conservatism coefficient of 0.85 after calculation. The conservatism coefficient reflects the uncertainty of the applicability of the population statistical record at the individual level. When a patient's postural activity pattern differs significantly from the population reference sample, the population statistical critical record may overestimate or underestimate the patient's actual critical approach ability. The reduction coefficient uniformly corrects this uncertainty.

[0037] The critical convergence values ​​are graded to generate contribution grading results. The grading is based on the distribution of critical convergence values ​​for all shift risk characteristic factors, using quantile segmentation to determine the boundaries of each level. Quantile segmentation adapts to the overall high or low distribution of critical convergence values ​​among different patients, avoiding the situation where a large number of factors concentrate in the same level under extreme distributions with a fixed threshold. Factors with negative critical convergence values ​​indicate that the shift risk tends to decrease with catheterization duration and are directly classified into the low contribution level, not participating in quantile ranking. Factors with non-negative critical convergence values ​​are categorized into each level using quantile segmentation: factors with critical convergence values ​​above the 75th percentile of non-negative factors are classified into the high contribution level; those between the 50th and 75th percentiles are classified into the medium-high contribution level; those between the 25th and 50th percentiles are classified into the medium-low contribution level; and those below the 25th percentile are classified into the low contribution level. This four-level grading, together with the low contribution level classification of negative factors, covers all shift risk characteristic factors, ensuring no factor is excluded from the grading range. The contribution grading results record the contribution level and corresponding critical convergence value of each shift risk characteristic factor. Multiple factors within the same contribution level are arranged in descending order of their critical convergence values. The order within a level determines the processing priority of factors at the same level during the hazard contribution level determination stage. For significantly shift-related factors, the contribution level is increased by one level based on the quantile division results in the contribution grading results. Significantly shift-related factors already at a high contribution level remain at that level without further increase. The increase reflects the additional hazard contribution confirmed by historical significant shift events. A factor whose critical convergence value falls within the medium-high contribution level may be increased to the high contribution level due to its association with a historical significant shift event. The contribution level of factors without historical confirmation is not increased, and their grading results strictly follow the quantile division results.

[0038] The risk contribution level is determined based on the contribution grading results. The contribution level of each factor in the contribution grading results, along with the absence of historical verification, jointly determines the final risk contribution level. Factors with high contribution levels and historical verification are classified as Level 1 risk contribution; high contribution levels without historical verification are classified as Level 2 risk contribution; medium-to-high contribution levels with historical verification are classified as Level 3 risk contribution; medium-to-high contribution levels without historical verification, as well as medium-to-low contribution levels, are uniformly classified as Level 4 risk contribution; and low contribution levels are classified as Level 5 risk contribution. Levels 1 to 5 correspond to decreasing priority in intervention resource allocation; higher-level factors receive more nursing resources in the risk intervention plan, and the greater the difference in level between factors, the more significant the difference in resource allocation. When the number of Level 1 risk contribution factors among the displacement risk characteristic factors of the same patient exceeds three, an overall risk upgrade label is triggered. This overall risk upgrade label indicates that the patient currently has multiple high-risk factors coexisting. Taking a patient with prolonged postoperative bed rest and sedation as an example, the consciousness impairment factor maintains a consistently high critical approach value during sedation; the loss of postural restraint factor simultaneously reaches a high contribution level due to the disappearance of active resistance under sedation; and the sedation drug use factor simultaneously enters a high contribution level due to the prolonged duration of sedation causing the disordered rhythm indicator depth to remain at the highest level. When these three factors concurrently reach Level 1 risk contribution within the same assessment period, their driving effect on catheter displacement exceeds the linear superposition of single-factor assessments, thus triggering an overall risk upgrade label. The risk contribution level is reviewed synchronously with the offset critical record update cycle. When a new critical offset event occurs, the critical approach value of the corresponding factor is recalculated, and the contribution classification result is updated accordingly.

[0039] In some embodiments, the step of generating a risk intervention plan based on the risk contribution level by performing factor weight difference mapping includes: screening high contribution factor items based on the risk contribution level; querying the cumulative intervention priority of the high contribution factor items across nursing cycles to obtain usable intervention configurations; generating a factor weight update table by performing weight update mapping based on the usable intervention configurations; and generating a risk intervention plan by performing nursing resource constraint matching based on the factor weight update table.

[0040] High-contribution factors are selected based on risk contribution levels. The selection criteria primarily focus on Level 1 and Level 2 risk contribution factors. Level 1 factors are directly included in the high-contribution factor list. Level 2 factors are simultaneously included when the overall confidence level of the shift risk characteristic factors is higher than 0.6; Level 2 factors with an overall confidence level lower than 0.6 are placed in the candidate observation list but not included in the main selection results. When overall risk escalation is triggered, the two highest-confidence Level 3 risk contribution factors are also included in the high-contribution factor list to ensure sufficient intervention coverage under overall high-risk conditions. When a patient has multiple Level 1 factors triggering escalation, some Level 3 factors may also have high practical intervention value under synergistic mechanisms, causing the actual risk contribution of Level 3 factors to exceed their level labeling range. High-contribution factors are arranged in descending order by risk contribution level as the primary key and overall confidence level as the secondary key. Level 1 factors are always ranked before Level 2 factors, and factors with higher overall confidence levels within the same level are ranked higher. This double-key ranking ensures that intervention resources are accurately allocated according to the actual risk contribution level in the subsequent weight mapping stage. Secondary factors in the candidate observation list are reassessed every 24 hours during the patient's hospitalization. When the confidence level rises to 0.6 or higher, they are automatically upgraded to be included in the main body of high contribution factor items. Candidate factors whose confidence level remains below 0.6 for more than 72 hours are downgraded to long-term observation labels. Long-term observation label factors no longer participate in daily review, but are re-triggered when a new offset event is added to the catheter historical positioning offset record.

[0041] For example, the step of querying the cumulative intervention priority across nursing cycles for the high-contribution factor item to obtain a usable intervention configuration includes: parsing the intervention priority hierarchy structure of the high-contribution factor item to generate a factor hierarchy relationship; locating a reference intervention template based on the factor hierarchy relationship to obtain a reference intervention template; performing backtracking verification of the displacement event incidence rate on the reference intervention template to obtain a valid template record; and performing adaptability screening based on the valid template record to form a usable intervention configuration.

[0042] Intervention priority hierarchy structure analysis was performed on high-contribution factors to generate factor hierarchical relationships. Within the same factor category, multiple factors are arranged vertically according to their risk contribution level. Different factor categories are prioritized horizontally according to their direct impact on catheter displacement, with factors directly affecting catheter tension changes having higher priority than those indirectly affecting catheter stability. The factor hierarchy is expressed as a directed graph, where nodes represent the factor entries within the high-contribution factors. Directed edges pointing from lower-level factors to higher-level factors indicate intervention dependencies. These dependencies suggest that interventions at lower-level factors achieve their maximum effect only after interventions at higher-level factors are completed. When a patient simultaneously has both a factor related to the degree of consciousness impairment and a factor indicating a lack of postural restraint, the actual immobilization effect of restraint measures strongly depends on the patient's cooperation. Uncontrolled consciousness impairment significantly weakens the restraint effect through the patient's active resistance. Therefore, the intervention edge for the lack of postural restraint points to the factor related to the degree of consciousness impairment, making the consciousness impairment factor the root factor in the intervention chain for this patient. In the factor hierarchy, nodes with an in-degree of zero are defined as root factors. Root factors are the starting point of the intervention chain, and their intervention priority receives the highest weight during the reference intervention template location phase. All factors in the high-contribution factor items are included in the factor hierarchy; no factor is excluded from the hierarchical structure. After hierarchical resolution, the directed graph of the factor hierarchy is checked for cycles. Cycles indicate that two factors are interdependent, forming a circular intervention deadlock. When a cycle is detected, the factors involved in the cycle are located in the high-contribution factor items, and the edges containing factors with lower overall confidence are deleted to eliminate the cycle, ensuring that the intervention chain has a clear start and end point.

[0043] Reference intervention templates are obtained by locating them based on the factor hierarchy. The combination of the root factor's factor category label and risk contribution level constitutes the core condition for template retrieval. Template entries matching this combination are searched in the historical nursing intervention template database. Each template record specifies the applicable factor category combination to define the template's coverage, includes an intervention measure list to detail specific nursing procedures, and indicates the applicable nursing cycle range to constrain the template's effective time period. Non-root factors in the factor hierarchy participate in template refinement as additional conditions after the root factor template is determined. The category labels of non-root factors further narrow down the range of candidate templates. The refined set of candidate templates covers the intervention needs of both root factors and all non-root factors. Candidate templates that cannot simultaneously cover all factors are excluded from the set. The reference intervention template is selected from the refined set of candidate templates, choosing the template entry with the most covered factors. If the number of covered factors is the same, the template whose applicable nursing cycle range best matches the current patient's catheterization duration is selected. The degree of matching in catheterization duration is measured by the proportion of the current catheterization duration falling within the template's applicable nursing cycle range; a higher proportion indicates a better match. When the overall risk upgrade label is activated in the high contribution factor item, the reference intervention template positioning requirement is that the candidate template must include synergistic intervention measures for the concurrent state of multiple factors in the factor hierarchy. Templates without synergistic measures will not be included in the reference intervention template even if they cover the most factors. When synergistic measures are missing, the suboptimal coverage template containing synergistic measures will be retrieved from the historical nursing intervention template library to replace it.

[0044] Valid template records were obtained by retrospectively validating the incidence of catheter displacement events using the reference intervention template. Retrospective data was derived from correlation analysis of hospital historical nursing records and catheter positioning deviation records. The retrospective validation window for the reference intervention template was set to all historical application records within the applicable nursing cycle. Templates with fewer than 5 historical application records were labeled as low-sample templates, and the retrospective validation results of low-sample templates were marked with a low-confidence label. The formula for calculating the incidence of displacement events is R = N_event / N_total, where N_event is the number of cases with catheter displacement events in historical application records, and N_total is the total number of cases in which the template was used historically. A lower R indicates stronger historical validity of the template. In the reference intervention templates, templates with an R-value below 0.15 are considered high-efficiency templates, those with an R-value between 0.15 and 0.30 are considered effective templates, and those with an R-value above 0.30 are marked as ineffective templates. Taking an intervention template designed for patients with altered consciousness and postural instability as an example, this template, in its historical application, required nursing staff to perform postural checks every two hours and adjust the tension of the catheter fixation strap. Retrospective records showed that the incidence of catheter displacement in patients using this template was significantly lower than in similar patients who did not use the template, and the R-value fell within the high-efficiency template range. Its historical effectiveness was fully supported in retrospective validation. High-efficiency templates and effective templates together constitute the effective template record. Ineffective templates are excluded from the effective template record. Low-sample, low-confidence labeled templates with an R-value below 0.15 are retained, and a low-sample inheritance label is added to the effective template record. Low-sample inheritance labeled templates need to undergo additional verification of current patient characteristics in subsequent suitability screening. Only after passing the similarity verification can they be included in the usable intervention configuration.

[0045] Adaptability screening is conducted based on valid template records to form usable intervention configurations. Adaptability screening precisely matches the applicable nursing cycle range of each template item in the valid template records with the current patient's nursing cycle stage. Complete coverage of the remaining duration of the current nursing cycle is a basic condition for template selection; templates covering only a portion of the remaining duration are marked as partially adapted. Partially adapted templates have additional time gap annotations in the usable intervention configuration to indicate the uncovered time range. High-efficiency templates from the valid template records are given priority in adaptability screening. Valid templates are used as a supplement when high-efficiency templates cannot provide a complete fit. Low-sample inherited annotation templates only participate in screening when neither high-efficiency nor valid templates meet the adaptation criteria. This screening hierarchy ensures that usable intervention configurations are primarily based on templates with the strongest historical effectiveness. The reusable intervention configurations are based on template items that have passed the suitability screening. Specific nursing procedures are determined using the intervention measure list within these template items. The measures in the intervention measure list are arranged in order of priority according to the root factor in the factor hierarchy, with measures corresponding to the root factor placed at the beginning of the list to ensure the intervention chain starts from the highest level. Measures corresponding to non-root factors follow sequentially in hierarchical order, maintaining consistency with the intervention dependency relationship to avoid execution reversal in the intervention chain. For reusable intervention configurations marked with time gaps, specific intervention parameters for the gap period are supplemented during the weight update mapping phase. These supplemented parameters are determined by referring to the statistical intervention intensity of similar patients during the gap period.

[0046] A factor weight update table is generated based on the available intervention configuration and weight update mapping. This mapping establishes a correspondence between the execution intensity parameters of each intervention in the available intervention configuration and the risk contribution level of each factor in the high-contribution factor category. Higher-level factors correspond to greater execution intensity parameters for their interventions. These execution intensity parameters are expressed as dimensionless weight coefficients; higher coefficients require nurses to invest more attention and effort in that intervention. The factor weight update table is built upon the available intervention configuration with current patient-specific adjustments. These adjustments are derived from historical shift patterns of significantly shifted factors in the shift risk characteristic factors. Historical shift patterns show that when a factor repeatedly triggers shifts after specific postural changes, the corresponding intervention's weight coefficient is additionally increased in the factor weight update table. The magnitude of this increase is determined by the normalized value of historical trigger frequencies. When the overall risk upgrade label is activated, the weight coefficients of all factors in the factor weight update table are multiplied by a non-linear enhancement coefficient of 1.2. This non-linear enhancement reflects the actual situation where intervention resource requirements exceed linear superposition under multi-factor concurrency. When three primary factors are activated simultaneously, the overall risk is not the sum of the three but exhibits a synergistic amplification effect. In the factor weight update table, the weight coefficients of the measures corresponding to the inherited annotation items of low samples are multiplied by a conservative reduction factor of 0.9 after integration. The weight coefficients after reduction still participate in the matching of nursing resource constraints. The conservative reduction reflects the objective limitation that there is uncertainty in the historical representativeness of low sample templates.

[0047] Risk intervention plans are generated by matching nursing resource constraints based on the factor weight update table. This matching process assigns the execution intensity parameters of each intervention measure in the factor weight update table to currently available nursing resources. The availability of each type of resource is predetermined within the current nursing cycle, and the total resource amount constitutes the upper limit constraint for generating intervention plans. When patients have multiple high-weight intervention needs simultaneously, the competition for similar resources among different intervention measures becomes particularly prominent. Bottlenecks in nursing staff scheduling, occupancy of follow-up equipment, and waiting times for bedside monitoring devices can all constitute actual implementation bottlenecks. The intervention measure with the highest weight coefficient in the factor weight update table is given priority in resource allocation. When weight coefficients are the same, resource supply is prioritized for the measure corresponding to the root factor according to the hierarchical order in the factor hierarchy. When the total resource demand of all intervention measures exceeds the upper limit of available nursing resources, the measures with the lowest weight coefficients are downgraded to a recommended implementation status rather than mandatory implementation. These downgraded measures are marked as resource-constrained downgrade items in the risk intervention plan. The actual implementation of these downgraded items depends on the flexible resource allocation by nursing staff during their shifts. If a patient's need for increased follow-up frequency conflicts with the follow-up time of other high-risk patients, the patient's follow-up interval is appropriately extended and marked as resource-constrained downgrade in the risk intervention plan. Simultaneously, alternative observation measures are added to partially compensate for the impact of insufficient follow-up frequency. The risk intervention plan specifies execution time nodes for each intervention measure to anchor the timing of operations, sets execution intensity parameters to quantify the required operational strength, and assigns execution status labels to distinguish between mandatory implementation, recommended implementation, and resource-constrained downgrade. The risk intervention plan is updated synchronously with the risk contribution level review cycle. When the level changes, the execution intensity parameters and time nodes of the corresponding intervention measures are rematched according to the new level.

[0048] Step S130: Extract scoring indicators from the risk intervention plan to form a shift scoring rule base, use the shift scoring rule base to identify a set of sensitive shift parameters, and perform benchmark archiving and storage based on the set of sensitive shift parameters to generate a shift risk file.

[0049] Specifically, a displacement scoring rule base is formed by extracting scoring indicators from the risk intervention program. Each intervention item in the risk intervention program corresponds to specific clinical observation indicators. These indicators need to be continuously monitored during the intervention to evaluate the intervention effect. Systematically extracting them constitutes the scoring dimensions for catheter displacement risk. Position management measures correspond to indicators such as the frequency of positional changes and the duration of rest, while catheter fixation measures correspond to indicators such as signs at the fixation site and catheter tension fluctuations. Clinical changes corresponding to missing dimensions form blind spots in the scoring system. Observation indicators corresponding to interventions in the risk intervention program that are enforced are given priority for inclusion in the displacement scoring rule base. Indicators corresponding to recommended implementation and resource-constrained downgraded items are included as supplementary items. Supplementary items are accompanied by secondary scoring labels to distinguish the scoring weight levels. The shift scoring rule base is indexed by scoring indicator name, specifying how to convert measured values ​​into standardized scores of 0 to 10 for each indicator. A brief interruption to the equipment when nursing staff change bed sheets or move patients can cause a sudden increase in the positional change frequency reading. This reading clearly originates from operational interference rather than actual positional change. This triggers anomaly labeling, which is not included in the current scoring summary. The anomaly label is simultaneously recorded in the patient's file for nursing staff review. Supplementary scoring indicators corresponding to resource-constrained downgraded items in the risk intervention plan are marked as weakly weighted indicators in the shift scoring rule base. The contribution weight of weakly weighted indicators is only 50% of that of mandatory indicators. If a patient's image follow-up examination is forced to extend the interval due to a time conflict, the corresponding indicator's acquisition frequency decreases and signal continuity is interrupted. Reducing the weight accurately reflects the decrease in the reliability of this indicator. Once the resource constraint is lifted, it will return to the standard level.

[0050] A set of sensitive displacement parameters was identified using a displacement scoring rule base. The identification process considered both the historical score fluctuation amplitude of each scoring indicator in the displacement scoring rule base and its temporal correlation with the displacement event. Indicators with larger fluctuation amplitudes and stronger temporal correlations with displacement events were considered more sensitive. Indicators with large fluctuation amplitudes but no significant temporal correlation with displacement events were not identified as sensitive parameters, thus avoiding misidentification of numerical fluctuations caused by environmental noise as displacement-sensitive signals. Cross-correlation analysis was performed on the scoring time series of each scoring indicator in the displacement scoring rule base within the most recent nursing cycle and the timestamps of displacement events in the historical catheter positioning and displacement records. The cross-correlation coefficient was calculated using the formula... The calculation is performed, where f(t) is the time series of the scoring indicator, E(t+τ) is the offset event indicator sequence, with a value of 1 at the time of the offset event and 0 at other times, and τ is the lead time. Indicators with a cross-correlation coefficient exceeding 0.6 and a lead time between 0 and 24 hours are identified as candidates for the sensitive shift parameter set. The lead time reflects the amount of time that the change in the indicator score precedes the occurrence of the offset event; the larger the lead time, the wider the warning window of the indicator. A patient's postural change frequency score showed a significant decrease approximately 12 hours before the occurrence of the offset event. Cross-correlation analysis identified that this indicator had a warning lead time of approximately 12 hours, and this indicator was included in the sensitive shift parameter set. The warning lead time information was simultaneously added to the corresponding indicator entry. The threshold for the cross-correlation coefficient of weakly weighted indicators in the sensitive shift parameter set has been increased to 0.7 to ensure that the sensitivity standard of the supplementary indicators is more stringent. New indicators without historical scoring time series in the shift scoring rule base are replaced by statistics of similar patient groups. The replacement data is labeled as the source of group statistics. The indicators from the source of group statistics participate in the scoring based on group representativeness rather than individual specificity. The deviation judgment adopts a more conservative threshold standard to deal with the risk of baseline shift caused by individual differences.

[0051] In some embodiments, the step of generating a displacement risk profile by performing baseline archiving and storage based on the sensitive displacement parameter set includes: performing individual baseline deviation comparison on the sensitive displacement parameter set to extract a subset of differential parameters; classifying the differential parameters according to the subset of differential parameters to generate a group of differential parameters; tracing historical records of the group of differential parameters to construct a parameter tracing set; and labeling patient information with risk prediction based on the parameter tracing set to form a displacement risk profile.

[0052] Individual baseline deviations of the sensitive displacement parameter set were compared to extract a subset of differential parameters. The individual baseline was determined by the mean scores of each sensitive displacement parameter set index within the first 48 hours after the patient's admission. In the early stage of admission, the catheterization procedure has not yet caused obvious postural stress response, and the mean scores during this period can accurately represent the individual's static baseline level. Due to differences in body type, activity habits, and basic physical signs, the static baseline values ​​of different patients on the same index may vary significantly. Using the patient's own measured baseline as a comparison reference can effectively eliminate the interference of such inter-individual systematic differences. The deviation of the current score of each indicator from the individual baseline is calculated using the formula Di = (S_i - μ_i) / σ_i, where S_i is the current score of the indicator, μ_i is the individual baseline mean, and σ_i is the baseline standard deviation. When σ_i is lower than a preset minimum value, the baseline standard deviation of the indicator in the same patient group is used instead to avoid division by zero. When the absolute value of Di_i exceeds 1.5, it is considered to deviate from the baseline. Indicators that deviate from the baseline are extracted as members of the difference parameter subset. A negative Di_i value indicates that the score of the indicator is lower than the individual baseline. A score lower than the baseline usually reflects a weakening of the actual control effect of the corresponding intervention. The deviation judgment threshold for population statistical source indicators in the difference parameter subset is raised to 2 standard deviations. The stress sensitivity values ​​of patients with a certain body type are systematically high. When the population mean is used as the baseline, the deviation will be overestimated. Raising the threshold can effectively suppress the false extraction caused by such systematic bias. For indicators in the sensitive shift parameter set that are missing data within the first 48 hours of admission, individual baselines cannot be established. During the differential parameter subset extraction stage, the corresponding indicators are replaced by the baseline of the same patient group and a baseline missing label is added. Indicators with missing baseline labels are classified into an independent low-confidence group. The overall deviation judgment of the low-confidence group is used in the differential parameter classification with reduced weight and is not merged with the normal baseline indicator conclusion. After the indicator has accumulated enough individual data, it will automatically switch back to the individual baseline mode.

[0053] Differential parameters are categorized into groups based on subsets of differential parameters. The categorization is based on the combination of the risk factor category to which each differential parameter belongs and the direction of deviation. Differential parameters of the same category and with consistent deviation directions are grouped together. Consistent directions indicate that multiple indicators under that risk factor category show a synergistic deterioration trend. Synergistic deterioration has a stronger driving force on the final displacement risk than a single deviation. When multiple scores related to postural constraints decrease synchronously, they are grouped into the same differential parameter group. The more members in a group, the wider the overall deterioration range of that risk category. Indicators in the differential parameter subset whose deviation exceeds three times the baseline standard deviation are marked as strong deviation parameters. Strong deviation parameters are listed separately within the differential parameter group as representative parameters of that group. Representative parameters serve as the starting point for historical record retrieval, and their retrieval results contribute a higher weight to the overall risk assessment of the group than ordinary member parameters. Differential parameters with missing baseline markers are grouped into a separate low-confidence group instead of being merged with normal parameters. The low-confidence group maintains an overall low-confidence label for subsequent retrieval, and its retrieval conclusions contribute a uniformly reduced weight to the final risk prediction. The difference parameter group records the list of member indicators, the average deviation within the group, and the representative parameter identifier for each group. The higher the average deviation, the greater the overall risk indication strength of the group. The average deviation and the number of members jointly determine the tracing retrieval priority of the difference parameter group. Groups with higher priority are retrieved first during tracing construction, and groups with lower priority follow in turn.

[0054] Historical record tracing is performed on the differential parameter groups to construct a parameter tracing set. The tracing path starts from the representative parameter of the differential parameter group and retrieves catheter displacement events that occurred when the historical score of the representative parameter showed similar deviations in the catheter historical positioning offset records. Similar deviations are defined as historical scores deviating in the same direction as the current deviation and the difference in deviation magnitude is within one time of the baseline standard deviation. Dual constraints ensure that the retrieved historical deviation events are comparable to the current deviation state in both direction and magnitude. The tracing scope of groups with more member indicators is expanded accordingly. The expansion method is to supplement the representative parameter tracing results with the independent tracing results of each member indicator. The union of the independent tracing results and the representative parameter tracing results forms the complete tracing record of the group. Records of repeated deviation events are merged and counted. The higher the count, the stronger the historical correlation between the deviation event and the current differential parameter group characteristics. The parameter traceability set consists of complete traceability records for all differential parameter groups. Each traceability record is associated with a corresponding differential parameter group identifier to locate the source of risk. An attached list of traceability deviation events is provided to record historical co-occurrence evidence. The strongest co-occurrence density of the group's features and historical deviation events is quantified by the highest correlation count. The traceability records of low-confidence groups are marked with an overall low-confidence label in the parameter traceability set. Differential parameter groups with a highest correlation count of zero indicate that similar deviations accompanied by deviation events have never occurred in history. This situation often occurs when the number of historical records is insufficient for newly introduced catheter types or rare clinical condition combinations. The corresponding traceability record is placed as an empty traceability record. The risk prediction is determined by referring to the average displacement probability obtained from the statistical analysis of the historical displacement incidence rate of patients with the same catheter type in the same ward within a similar catheterization duration interval.

[0055] Patient information is associated with a parameter traceability set to form a displacement risk profile. The associated patient information dimensions include current catheterization duration, past displacement history, and current nursing level. These three pieces of information are cross-referenced with the historical offset patterns of each traceability record in the parameter traceability set. Patients with a high degree of overlap between catheterization duration and historical high-incidence periods, a past displacement history, and a lower current nursing level are marked with a high-risk pre-judgment label. The higher the highest association count of each traceability record in the parameter traceability set, the stronger the confidence level for triggering a high-risk label after combining it with patient information. If a patient's current differential parameter group features show a high-count association in historical traceability, and the patient has a history of two catheter displacements, the cross-comparison determines a high-confidence, high-risk profile, and the pre-judgment label level is correspondingly upgraded to the highest level. For empty traceability records in the parameter traceability set, the pre-judgment label level is determined by referring to the population statistical displacement probability after associating the patient information. If the population statistical probability exceeds the current patient ward average, it is labeled as medium risk; if it is below the average, it is labeled as low risk. These two risk levels correspond to different levels of intervention response intensity in monitoring resource allocation decisions. The displacement risk profile consists of a summary of risk prediction labels corresponding to all differential parameter groups. Each entry includes three parts: the differential parameter group identifier, the risk prediction label level, and the basis for association with patient information. The prediction level of low-confidence label entries is downgraded by one level in the final displacement risk profile. High-risk labels in the low-confidence state are downgraded to medium-risk to avoid over-warning based on unreliable data. The downgrade result is accompanied by an explanation of the downgrade reason, recording the triggering source of the original high-risk label and the low-confidence attribute on which the downgrade judgment is based, so that the risk profile can reflect the current confidence state while retaining a complete label history.

[0056] Step S140: Perform scoring item association analysis on the displacement risk file to extract the displacement risk association chain, extract time-series drift features based on the displacement risk association chain to generate trend-weighted labels, and arrange the trend-weighted labels according to the risk priority to form a graded displacement assessment sequence.

[0057] Specifically, a correlation analysis of scoring items is performed on the displacement risk profile to extract the displacement risk correlation chain. This analysis maps the causal and concurrent relationships between the scoring indicators involved in each risk prediction labeling item in the displacement risk profile—causal relationships meaning that the deterioration of one indicator precedes the deterioration of another in time, and concurrent relationships meaning that two indicators historically tend to deviate simultaneously—to an ordered chain structure. These two types of relationships jointly determine the edge type and direction of the chain structure. The differential parameter group member indicators corresponding to each item in the displacement risk profile are constructed into directed chains according to their historical deviation time-series relationships. These time-series relationships are determined by statistically analyzing the order in which the scores of each indicator change before each offset event occurs in the parameter tracing set. Indicators that deviate first are located at the front of the chain, and indicators that deviate later are located at the back of the chain. Concurrent indicators with unclear time-series relationships are placed side-by-side at the same chain node. The head node of the displacement risk association chain corresponds to the scoring indicator that first shows a worsening signal. This node is the starting point of the entire risk transmission path. In the displacement risk transmission path of a patient who has been bedridden for a long time after surgery, the deterioration of the positional change frequency score always occurs earlier than the deterioration of the catheter fixation tension score. The former's persistent low-frequency signal is the initial alarm of the entire transmission chain. Identifying the head node allows the risk transmission structure to be clearly presented at the file level. When the differential parameter group corresponding to the low-confidence labeled items in the displacement risk file participates in the chain construction, the weight of the association edge between its member indicators and the member indicators of other normal-confidence items is multiplied by a reduction factor. After reduction, the association edge with a lower weight is ranked later in the chain analysis.

[0058] In some embodiments, the step of extracting time-series drift features and generating trend-weighted identifiers based on the shift risk association chain includes: extracting time-series score records for each score node based on the shift risk association chain; calculating the drift slope of the time-series score records to obtain the score drift rate; performing trend intensity classification based on cumulative drift according to the score drift rate to generate trend intensity identifiers; and performing weighted coefficient mapping according to the trend intensity identifiers to generate trend-weighted identifiers.

[0059] The temporal score records for each scoring node are extracted based on the displacement risk association chain. All score values ​​of the corresponding scoring indicators for each node in the displacement risk association chain within the most recent nursing cycle are arranged in ascending order by the record timestamp; the arrangement result is the temporal score record for that node. The temporal score records of the indicators corresponding to the first node in the displacement risk association chain are extracted first, as early changes in the score of the first node directly determine the rate of risk transmission to subsequent nodes. In patients who have been under long-term sedation, the positional change frequency score of the first node usually increases first during the recovery period after sedation medication adjustment. This signal predicts that subsequent chain node scores, such as catheter tension, are likely to improve. The completeness of the temporal score records of the first node plays a fundamental role in the quality of identifying the drift trend of the entire chain. Any long-term missing period in the first record will create a chain-like analytical gap at the transmission chain level. In time-series scoring records, time periods with missing scores are filled using linear interpolation of two adjacent valid scores. The filled values ​​are labeled with interpolation markers to distinguish them from the measured values. Time periods with consecutive missing scores exceeding 4 hours are not interpolated and are retained as missing segments. Missing segments are segmented during the drift slope calculation phase, with slopes calculated independently on both sides of each missing segment without merging across segments. Scores marked as data anomalies in time-series scoring records are not included in the time-series arrangement. The time position of the anomaly is replaced by the mean of the valid values ​​before and after it, with a replacement marker added. Interpolation markers and replacement marker segments participate in the calculation with reduced weights within their respective windows.

[0060] For example, the step of calculating the drift slope of the time-series scoring records to obtain the scoring drift rate includes: statistically analyzing the scoring node intervals of the time-series scoring records to generate interval distribution features; identifying abrupt score change drift segments based on the interval distribution features to generate acceleration segment identifiers; performing slope fitting analysis on the acceleration segment identifiers to form slope fitting parameters; and calculating the scoring drift rate based on the slope fitting parameters by locating segmented slope breakpoints.

[0061] Interval distribution features are generated by statistically analyzing the intervals between scoring nodes in time-series scoring records. The scoring node interval is the difference between the timestamps of two adjacent valid scoring records. The interval values ​​of all adjacent valid scoring nodes in the time-series scoring records are arranged chronologically. The mean, standard deviation, and maximum value of the interval sequence are extracted to constitute the interval distribution features. The mean reflects the average monitoring frequency, the standard deviation reflects the stability of the monitoring frequency, and the maximum value captures the longest monitoring gap period. These three statistics jointly describe the temporal completeness of the scoring indicator. An interval distribution feature with a standard deviation exceeding 50% of the mean is considered a high-variability interval distribution. High-variability distributions indicate unstable monitoring frequencies. For example, in a scenario where night shift nursing staff are reduced, the scoring monitoring interval for a patient during the night shift can be 3 to 4 times longer than that during the day shift. The alternation of day and night shift interval values ​​causes the standard deviation of the interval sequence to be significantly larger. If a uniform threshold is used to judge sudden changes without differentiation, the difference between the last score at night and the first score of the next morning shift is easily misjudged as a sudden change. The high-variability judgment mechanism is specifically designed to address such scenarios with uneven monitoring rhythms. The node intervals corresponding to the interpolation-filled segments in the time-series scoring records are not included in the statistical calculation of the interval distribution characteristics. The interpolation nodes are virtual times rather than actual monitoring times. Including them would artificially compress the interval mean, resulting in a lower threshold for judging high variation.

[0062] Based on the interval distribution characteristics, abrupt score shifts are identified, generating accelerated segment identifiers. The essential characteristic of an abrupt shift segment is a jump in score within a short period of time that significantly exceeds the normal fluctuation range, rather than a slow drift. Segments where the score difference between adjacent score nodes exceeds twice the individual baseline standard deviation for that indicator are identified as abrupt shift segments. When the interval distribution characteristics are highly variable, the judgment threshold is raised. Taking the scenario of a patient being transferred to the intensive care unit after surgery as an example, the time interval between the first score record after transfer and the last record before transfer can exceed 6 hours. During this gap, the patient's positioning and management environment have undergone fundamental changes. The difference between the two scores is actually the cumulative result of multiple gradual steps rather than a single acute jump. Raising the high variability threshold prevents such long-interval cross-environment differences from being mistakenly included in the abrupt shift segment, thus distinguishing it from true short-term acute fluctuations. The acceleration segment identifier records the start and end timestamps of each abrupt change drift segment, along with the total score change within the segment. A larger total score change indicates a more significant drift amplitude in that abrupt change segment. The degree of temporal overlap between the abrupt change segment and the abrupt change period of the score at the head node of the chain associated with the shift risk is recorded in the acceleration segment identifier. Higher temporal overlap indicates that the risk transmission at the head node is in a concentrated phase. The length of the time-series score record is insufficient to support the identification of abrupt change segments; the time interval distribution characteristics are only used as an auxiliary reference. The acceleration segment identifier is an empty set, and the corresponding slope fitting analysis performs an overall fit on the entire sequence without distinguishing between acceleration segments.

[0063] Slope fitting analysis was performed on the accelerated zone markers to generate slope fitting parameters. The slope fitting analysis used each abrupt change segment within the accelerated zone marker as a boundary, dividing the time-series score records into three parts: the pre-abrupt change segment, the abrupt change segment, and the post-abrupt change segment. Linear regression was performed independently on each part. Independent segmental fitting avoids the influence of extreme values ​​from the abrupt change segment on the slope estimation of non-abrupt change periods, making the drift trend representation of each time period more accurately reflect actual clinical changes. Score changes within abrupt change segments are drastic and the time span is usually short, resulting in a wider confidence interval for the linear regression slope. High uncertainty annotations were added to the slope fitting parameters for abrupt change segments. These high uncertainty annotations trigger conservative weighting of the slope for that segment during the segmented slope breakpoint localization stage. The slope fitting parameters record three items for each segment: the regression slope, the coefficient of determination R², and the number of effective sample points. A higher R² indicates a better fit of the linear regression to the score changes in that segment. Segments with an R² below 0.4 and fewer than 5 effective sample points are marked as having low fit quality. Low fit quality suggests that the score changes in that period do not show a linear trend but exhibit nonlinear fluctuations. For example, during a patient's rehabilitation training, nursing staff assisted in completing a round of active postural training every 2 hours. The frequency of postural changes in the score showed a periodic pattern of alternating peaks and troughs with the training rhythm. The linear regression R² was low during the training period, and this period was marked as nonlinear in the slope fitting parameters, with its slope estimate not participating in the linear extrapolation of the trend direction. When the acceleration segment is marked as an empty set, the slope fitting parameters are filled with the overall regression results of the entire sequence. The overall regression results also carry R² and the number of effective sample points.

[0064] The score drift rate is calculated by segmenting slope breakpoints based on slope fitting parameters. Breakpoint location identifies positions in the slope fitting parameters where the regression slope changes significantly between adjacent time periods. When the absolute value of the difference between two adjacent slopes, |Δβ_k| = |β_{k+1} - β_k|, exceeds a threshold, that position is marked as a drift breakpoint. The threshold is determined by the 75th percentile of the absolute value of the absolute slope across all segments for that patient, with units consistent with β_k (scores / hour). If the slope remains negative and its absolute value continues to increase after a breakpoint where the slope changes from positive to negative or |Δβ_k| suddenly increases, it is identified as an accelerated deterioration segment. The existence of an accelerated deterioration segment is one of the core conditions for triggering a high-risk level. The rate of score drift is defined as a piecewise time-varying rate function v(t) = β_k, t ∈ [t_k, t_{k+1}), with the drift breakpoint as the switching moment. The function value is always equal to the regression slope of each segment. At the breakpoint, the rate jumps. The entire function clearly presents the staged evolution of the score from stable, slow decline to accelerated deterioration. For example, on the first day of admission, the slope of the positional change frequency score of a patient with catheterization is close to zero. On the second night, |Δβ| exceeds the threshold, triggering the first breakpoint, and the slope turns negative. On the third day, the absolute value of the slope continues to expand, forming an accelerated deterioration segment. The three function values ​​and the two breakpoint moments together constitute the complete rate of score drift for this patient. In the slope fitting parameters, the slope corresponding to the sudden change segment is weighted down when locating the breakpoint. Short-term extreme changes in the sudden change segment can easily cause |Δβ_k| to be artificially high, resulting in false breakpoints. After reduction, the contribution of the slope of the sudden change segment to the breakpoint determination is reduced. The brief score jump caused by the nurse turning the patient will not be identified as a trend drift breakpoint. The timestamps of drift breakpoints mark critical moments of risk state transitions. Dense breakpoints indicate frequent changes in the patient's displacement risk state, while sparse breakpoints suggest a relatively stable risk state. In the slope fitting parameters, the score drift rate component corresponding to the low-fit quality labeled segment is given a low-confidence label, and the low-confidence component participates in the accumulation with a reduced weight in the trend strength calculation.

[0065] Trend intensity labels are generated by classifying trend intensity based on cumulative drift according to the rating drift rate. The generation of trend intensity labels first involves multiplying and summing the slopes of each segment of the rating drift rate with the corresponding time period length to obtain the cumulative drift amount. The calculation formula is L=Σ(β_k×Δt_k), where β_k is the regression slope of the k-th segment, Δt_k is the corresponding time period length, and the sign of L reflects the direction of cumulative offset. The larger the absolute value, the deeper the total deviation of the rating from the individual baseline within the observation period. When there is an accelerated deterioration segment in the score drift rate, the cumulative drift contribution corresponding to that segment is multiplied by an enhancement coefficient. In one case, a patient's postural change frequency plummeted from normal to near zero within just 4 hours after a sharp decline in nighttime consciousness. Simultaneously, the catheter fixation tension score declined due to passive postural accumulation. The negative drift density of both indicators in the accelerated deterioration segment far exceeded that in the previous uniform decline phase. The absolute value of β_k in this segment was already relatively large. Although Δt_k was short, its cumulative contribution after multiplying by the enhancement coefficient was significantly higher than that in the uniform decline segment of the same duration. This allowed the L value to accurately reflect the substantial increase in overall drift depth caused by this high-density deterioration phase. After the cumulative drift was calculated, the ratio of the absolute value of L to the individual baseline standard deviation was used as the grading criterion to determine the level of trend intensity. A ratio greater than 3 indicated high trend intensity, a ratio greater than 1.5 but less than 3 indicated medium trend intensity, and a ratio less than 1.5 indicated low trend intensity. The cumulative contribution of low-confidence label drift components is weighted down in the trend strength calculation. If the cumulative drift amount still reaches the high trend strength threshold after the reduction, it indicates that the risk trend is still significant even under conservative estimation. In this case, the trend strength label maintains a high trend strength judgment and does not downgrade.

[0066] Trend-weighted identifiers are generated by performing weighted coefficient mapping based on trend strength identifiers. This mapping converts the three strength levels of the trend strength identifiers into quantitative weights for hazard priority ranking, with higher trend strength corresponding to the highest weight coefficient and lower trend strength corresponding to the lowest. In the shift risk association chain, the trend strength identifier of the head node receives an additional chain position coefficient bonus during weighted coefficient mapping. The final weight of the head node is equal to the product of the base weight coefficient corresponding to the trend strength level and the chain position coefficient. The chain position coefficient of the head node is greater than 1, while the chain position coefficient of the middle node is equal to 1. When two nodes on the same transmission chain have the same trend strength, the head node receives a higher final weight due to the chain position coefficient bonus. This mechanism is particularly crucial in scenarios with multiple concurrent risk transmission chains—when both the body position change frequency chain and the duct tension chain exhibit medium trend strength, their respective head nodes are ranked before the middle nodes of their respective chains due to the chain position coefficient, ensuring that the starting point of each transmission chain preferentially enters the front position of the hazard ranking. The trend-weighted identifier records the identifier, trend strength identifier, and final weighting coefficient of each scoring node. These three elements together identify the node's identity, current risk intensity, and weight position in the ranking. Isolated nodes, lacking mutual verification through chain relationships, have their weighting coefficients reduced after mapping. After the trend-weighted identifier is generated, the weights of each node are verified for consistency according to the chain order of the shift risk association chain. The weight of the first node in the chain should not be lower than that of its subsequent nodes. If the weight of a subsequent node is consistently higher than that of the first node in two consecutive update cycles, it indicates that the transmission order of the score drift is inconsistent with historical patterns and is not due to normal time lag, triggering a re-verification of the topology of the shift risk association chain.

[0067] A graded shift assessment sequence is formed by prioritizing trend-weighted indicators according to risk level. The risk priority ranking uses the final weighted coefficient of the trend-weighted indicators of all scoring nodes as the primary key, with the node with the highest weighted coefficient placed first. If the weighted coefficients are the same, the intensity level of the trend indicator is used as the secondary key. If the intensity levels are the same, the first node of the chain is prioritized as the third-level ranking criterion. This three-key ranking ensures that nodes with high risk, high trend intensity, and at the starting point of risk transmission are always at the forefront of the graded shift assessment sequence. The graded shift assessment sequence is divided into three levels based on the arrangement results: high-risk, medium-risk, and low-risk. The high-risk group covers nodes with weighted coefficients higher than the mean plus one standard deviation of all nodes, while the low-risk group covers nodes lower than the mean minus 0.5 standard deviations. The asymmetric setting of using one standard deviation for the upper bound and 0.5 standard deviations for the lower bound narrows the medium-risk group towards high-risk to ensure that potential high-risk nodes are not missed. The medium-risk group covers the remaining nodes. The boundaries of the three groups are dynamically adjusted according to the weighted coefficient distribution of the current nursing cycle, not dependent on fixed thresholds. In the graded shift assessment sequence, the weighted coefficients of nodes in the high-risk group are higher than those in the medium- and low-risk groups. The difference in coefficients between groups is dynamically adjusted according to the trend intensity distribution of each node. When multiple chain head nodes in the high-risk group are simultaneously at a high trend intensity, the difference in coefficients between groups widens significantly, reflecting the superimposed situation of multiple risk transmission chains advancing concurrently. In the trend-weighted identification, nodes that trigger re-verification due to consistency verification of weighting coefficients are marked with a verification pending label in the graded shift assessment sequence. The nodes to be verified are temporarily moved to the end of the medium-risk group. After verification, the weighting coefficients are recalculated and the arrangement is updated based on the corrected shift risk association chain.

[0068] Step S150: Perform displacement probability calculation on the graded displacement assessment sequence to obtain displacement risk probability value; delineate the stratified monitoring range based on the fluctuation range of the displacement risk probability value to generate stratified early warning indicators; and perform re-examination frequency matching based on the stratified early warning indicators and risk factor set to generate catheter tip displacement risk assessment results.

[0069] Specifically, a displacement probability calculation is performed on the graded displacement assessment sequence to obtain the displacement risk probability value. The displacement probability calculation inputs the node scores and corresponding weighting coefficients of each hazard group in the graded displacement assessment sequence into a probability model. The model input is a weighted vector of the current node score and the weighting coefficients, and the output is a displacement risk probability estimate between 0 and 1. The displacement risk probability value P is given by the formula... The formula is determined as follows: N is the total number of scoring nodes in the graded shift assessment sequence, M is the total number of risk factor entries, s_i is the normalized value obtained by dividing the current score of the i-th scoring node by the full score of 10, ranging from 0 to 1, w_i is the trend weighting coefficient of the i-th node, h_j is the comprehensive confidence of the j-th factor, ranging from 0 to 1, α_j is the mean of the confidence of each matching in the factor association table for the j-th factor, ranging from 0 to 1, and sig is the Sigmoid function that maps the linear weighted sum to the probability interval of 0 to 1. The superposition of the two inputs ensures that the shift risk probability value is simultaneously constrained by both the scoring drift trend and the risk factor background. In the graded displacement assessment sequence, the trend-weighted coefficient w_i of the high-risk group nodes is multiplied by the intra-group enhancement coefficient before being substituted into the formula. The enhancement coefficient reflects the superlinear contribution of the high-risk node to the displacement probability. When the trend intensity of the first node in the high-risk group remains at a high level, each decrease in the node's score significantly increases the P-value more than the same decrease in the low-risk group. The inter-group coefficient differentiation ensures that the P-value responds to high-risk signals preferentially over low-risk signals. The weighted coefficients of the low-risk group nodes are retained in the calculation. The displacement risk probability value is calculated with the current time as the baseline time point, outputting the current probability estimate. The calculation is performed on a rolling basis within each nursing cycle as the score is updated, and the rolling calculation cycle is aligned with the update cycle of the trend-weighted indicator.

[0070] In some embodiments, the step of defining the stratified monitoring range and generating stratified early warning identifiers based on the fluctuation range of the displacement risk probability value includes: dividing the displacement risk probability value into segments according to the monitoring time window to generate a continuous time period probability sequence; performing fluctuation range statistics on the continuous time period probability sequence in relation to nursing operations to obtain the fluctuation interval distribution; allocating monitoring priorities according to the fluctuation interval distribution to generate a stratified patient list; and determining the stratified early warning identifier based on the stratified patient list.

[0071] The probability values ​​of displacement risk are segmented according to the monitoring time window to generate a continuous time-segment probability sequence. The length of the monitoring time window is set with reference to the nursing shift handover cycle. The probability estimates within the handover cycle are grouped into the same window. The mean and range of the probability values ​​within the window together describe the overall level and fluctuation of displacement risk within that handover period. The boundary of the handover cycle serves as a natural dividing point between adjacent windows, ensuring that the segmentation of the continuous time-segment probability sequence is synchronized with the rhythm of nursing work. The rolling calculation results of the displacement risk probability values ​​within the same window are arranged in chronological order to form the probability subsequence of that window. The length of the subsequence depends on the number of score updates within the window. Windows with higher update frequencies have longer subsequences and finer granularity in identifying drift trends. The continuous time-period probability series uses the probability mean to characterize the average level of displacement risk probability values ​​within each time window, and the probability range to characterize the fluctuation range of risk estimates within that time period. Simultaneously, the number of score updates within a window is recorded as a criterion for data sufficiency. For a certain patient, the fluctuation range of displacement risk probability values ​​during intensive daytime nursing procedures was significantly higher than that during quiet nighttime procedures, and the probability range of the former was significantly larger than that of the latter. These two characteristics form a clear time-period contrast in the continuous time-period probability series. When the number of score updates within a window is less than 3, the probability mean of that window is labeled with a low-sample annotation. The probability mean of the low-sample-annotated window is used in subsequent statistics with reduced weight.

[0072] The fluctuation range distribution was obtained by statistically analyzing the fluctuation amplitude of a continuous probability sequence in relation to nursing procedures. The fluctuation amplitude statistics correlated the probability range of each time window in the continuous probability sequence with the type and frequency of nursing procedures occurring within that window. Nursing procedure information was extracted from nursing execution records by timestamp matching. The operation type and frequency were summarized by window and then correlated with the probability range. Operation types with high correlation coefficients were identified as fluctuation-related operations. The more frequently a fluctuation-related operation occurred within the current window, the greater the expected fluctuation amplitude for that window. Different operation types exhibited significantly different degrees of disturbance to catheter position. Suctioning requires coordinated movement of the patient's head and neck, resulting in significant tension disturbance to the cervical catheter. Dressing changes, however, were limited to a fixed location, having a relatively limited impact on the overall catheter tension. When both types of operations occurred simultaneously in the same patient within the same window, their additive effect manifested in a synergistic increase in the probability range. The fluctuation interval distribution uses time windows as the horizontal axis and probability range as the vertical axis to describe the fluctuation distribution of each window. Continuous time windows with consistently high fluctuation amplitudes are marked as high-fluctuation periods. High-fluctuation periods and low-fluctuation periods form a visual partition feature in the fluctuation interval distribution, and the partition boundaries correspond to key moments of change in nursing operation density. The probability range of low-sample labeled windows in the continuous time probability sequence is weighted less in the fluctuation interval distribution statistics.

[0073] A stratified patient list is generated by prioritizing monitoring based on the distribution of fluctuation intervals. The monitoring priority allocation combines the coverage ratio of high-fluctuation periods for each patient within the fluctuation interval distribution with the mean probability range. Patients with higher coverage ratios and larger mean ranges are assigned higher monitoring priority. A high fluctuation coverage ratio indicates that the patient's displacement risk is unstable for most of the time, requiring continuous high-frequency monitoring to capture moments of risk abrupt change. For patients with clearly correlated fluctuations within the fluctuation interval distribution, the overlap between high-fluctuation periods and operation times is also included in the priority assessment. A high degree of overlap indicates that the patient's displacement risk fluctuations are mainly driven by predictable nursing operations, allowing for targeted enhanced monitoring before and after the operation. For example, if a patient's displacement risk probability only briefly increases after a turning operation and then quickly decreases, monitoring resources for this patient are concentrated before and after the turning operation window, while the monitoring intensity during resting periods unrelated to high-risk periods is correspondingly reduced, ensuring that limited nursing monitoring resources are fully utilized during periods when risk is actually concentrated. The stratified patient list ranks all patients with catheters in the current ward according to monitoring priority from highest to lowest. Each entry is indexed by a patient identifier, with priority levels indicating the urgency of monitoring. Fluctuation characteristics are used to describe the coverage percentage of high-fluctuation periods and the types of procedures associated with the main fluctuations for nursing staff reference. Priority levels are divided into three tiers: Level 1 (intensive monitoring), Level 2 (regular follow-up), and Level 3 (routine follow-up), each corresponding to different minimum follow-up frequency requirements. Patients with a high coverage percentage in a low-sample annotation window have their priority appropriately increased in the stratified patient list to conservatively address the possibility of underestimation of risk due to data sparsity; the increase is limited to no more than one priority level.

[0074] Stratified patient alert indicators are determined based on the stratified patient list. Each priority level in the stratified patient list directly maps to its corresponding alert indicator level: Level 1, close monitoring, corresponds to a red alert indicator; Level 2, regular follow-up, corresponds to a yellow alert indicator; and Level 3, routine follow-up, corresponds to a green alert indicator. This three-color system is consistent with commonly used clinical risk stratification color standards. In addition to the indicator level, each stratified alert indicator includes the core criteria that triggered it. These criteria include the coverage rate of high-fluctuation periods for the corresponding patient, the types of operations associated with the main fluctuations, and the number of high-risk group nodes in the graded shift assessment sequence. These three criteria help nurses quickly understand the logic behind the current alert, preventing hesitation in handling situations due to a lack of judgment when faced with alert levels. For example, the core criteria accompanying a patient's red alert indicator show that the coverage rate of high-fluctuation periods reaches 80% throughout the day and the number of high-risk group nodes is 4. Based on this, nurses judge that the patient is in a continuous high-risk state throughout the entire period, rather than experiencing operation-driven periodic fluctuations. The management strategy is then adjusted to continuous intensive monitoring rather than focused monitoring during operation windows. Patients whose priority was raised due to low sample annotation windows in the stratified patient list will have a "Data Needs Improvement" annotation added to the stratification warning label. This reminds nursing staff to proactively supplement the patient's score monitoring data to verify the accuracy of the upgraded priority. After the data is supplemented, the stratification warning label is regenerated based on the updated fluctuation range distribution. If the updated priority is consistent with the one before the upgrade, the "Data Needs Improvement" annotation is removed. If the priority drops, a downgrade label is simultaneously added, and the history of this upgrade is recorded for subsequent risk pattern analysis. After the stratification warning label is generated, a notification is pushed to the corresponding responsible nurse. The notification includes the patient identifier, warning level, and suggested follow-up time window. Red warning labels trigger immediate push notifications, while yellow and green warning labels are pushed in batches along with shift handover information.

[0075] The risk assessment of catheter tip displacement is generated by matching the re-examination frequency based on stratified warning indicators and risk factor sets. Re-examination frequency matching determines the imaging re-examination frequency by combining the level of the stratified warning indicator with the overall confidence level of each factor in the risk factor set. The stratified warning indicator level determines the lower limit of the re-examination frequency, while the number of high-confidence factors in the risk factor set determines the upward adjustment range of the re-examination frequency from the lower limit. Both factors jointly constrain the final re-examination frequency value. If a patient with a yellow warning has multiple concurrent high-confidence factors, the re-examination frequency is further tightened from the lower limit of the yellow warning, ensuring that the actual re-examination schedule matches the true risk level after the multi-factor aggregation. When the stratified warning indicator is red and the number of primary risk contributing factors in the risk factor set reaches the overall risk escalation threshold, the highest re-examination frequency is applied. Patients with persistently deepening consciousness impairment, complete failure of postural restraint, and persistently high-level disordered rhythm indicators due to sedation drug use simultaneously activate all three high-risk factors, triggering the highest re-examination frequency requirement. Candidate appendix items in the risk factor set are not included in the factor count during the follow-up frequency matching phase. Including candidate items with insufficient clinical confirmation would artificially inflate the follow-up frequency. The catheter tip displacement risk assessment results are indexed by patient identifiers, recording the current stratified warning level and matched follow-up frequency to guide monitoring arrangements. A core factor list supporting risk assessment is also included to help nurses understand the risk causes, and the recommended follow-up time is indicated to clarify the execution node. The core factor list extracts the top three items with the highest overall confidence from the risk factor set, allowing nurses to directly understand the main driving factors of the current risk when reviewing the results. The catheter tip displacement risk assessment results are updated synchronously with the rolling update cycle of displacement risk probability values. When the stratified warning level upgrades to red, it is detected and triggered immediately using an event-driven approach, independent of the rolling cycle, and an immediate push notification is generated. Changes in other levels are pushed synchronously with the rolling cycle.

[0076] To implement the catheter tip displacement risk assessment method corresponding to the above method embodiments, in order to achieve the corresponding functions and technical effects. See also Figure 2 , Figure 2 This diagram illustrates a structural block diagram of a catheter tip displacement risk assessment system 200 provided in an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The catheter tip displacement risk assessment system 200 provided in this embodiment includes:

[0077] Risk factor module 201 is used to collect patient position change frequency data and catheter historical positioning deviation records, and perform clinical risk factor extraction to form a risk factor set based on the rhythm disorder characteristics of the position change frequency data;

[0078] Risk assessment module 202 is used to extract displacement risk feature factors using the risk factor set and the catheter historical positioning offset record, perform critical displacement risk contribution assessment on the displacement risk feature factors to generate risk contribution level, and perform factor weight difference mapping based on the risk contribution level to generate risk intervention plan;

[0079] The archive storage module 203 is used to extract scoring indicators from the risk intervention plan to form a shift scoring rule library, use the shift scoring rule library to identify a set of sensitive shift parameters, and perform benchmark archiving and storage based on the set of sensitive shift parameters to generate a shift risk archive;

[0080] The trend arrangement module 204 is used to perform scoring item association parsing on the displacement risk file to extract displacement risk association chains, extract time-series drift features based on the displacement risk association chains to generate trend weighted identifiers, and arrange the trend weighted identifiers according to the risk priority to form a graded displacement assessment sequence.

[0081] The stratified early warning module 205 is used to perform displacement probability calculation on the stratified displacement assessment sequence to obtain displacement risk probability value, delineate the stratified monitoring range based on the fluctuation range of the displacement risk probability value to generate a stratified early warning identifier, and perform review frequency matching based on the stratified early warning identifier and the risk factor set to generate catheter tip displacement risk assessment result.

[0082] The catheter tip displacement risk assessment system 200 described above can implement one of the catheter tip displacement risk assessment methods in the above method embodiments. The options in the above method embodiments are also applicable to this embodiment, and will not be detailed here. The remaining content of this application embodiment can be referred to the content of the above method embodiments, and will not be repeated in this embodiment.

[0083] The purpose of the above embodiments is to reproduce and derive the technical solution of the present invention by way of example, and to fully describe the technical solution, purpose and effect of the present invention. The purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosure of the present invention, and not to limit the scope of protection of the present invention.

Claims

1. A method for assessing the risk of catheter tip displacement, characterized in that, include: Collect patient position change frequency data and catheter historical positioning deviation records, and perform clinical risk factor extraction to form a risk factor set based on the rhythm disorder characteristics of the position change frequency data; Using the risk factor set and the catheter historical positioning offset record, displacement risk feature factors are extracted. The critical displacement risk contribution of the displacement risk feature factors is evaluated to generate a risk contribution level. Based on the risk contribution level, factor weight difference mapping is performed to generate a risk intervention plan. The risk intervention plan is subjected to scoring indicators to form a shift scoring rule library. The shift scoring rule library is used to identify a set of sensitive shift parameters. Based on the set of sensitive shift parameters, a benchmark archiving and storage is performed to generate a shift risk profile. The displacement risk profile is analyzed by performing a scoring item association parsing to extract the displacement risk association chain. Based on the displacement risk association chain, the temporal drift features are extracted to generate a trend-weighted identifier. The trend-weighted identifiers are then arranged according to the risk priority to form a graded displacement assessment sequence. The displacement probability is calculated to obtain the displacement risk probability value of the graded displacement assessment sequence. The stratified monitoring range is defined based on the fluctuation range of the displacement risk probability value to generate a stratified early warning identifier. The re-examination frequency is matched based on the stratified early warning identifier and the risk factor set to generate the catheter tip displacement risk assessment result.

2. The method according to claim 1, characterized in that, The process of extracting clinical risk factors based on the rhythmic disturbance characteristics of the postural change frequency data to form a risk factor set includes: The rhythmic pattern analysis of the body position change frequency data is performed to obtain rhythmic feature sequences; Based on the rhythmic feature sequence, identify disordered rhythmic segments in a state of impaired consciousness and generate disordered rhythm identifiers; Based on the disordered rhythm identifiers, clinical risk factors are matched to form a factor association table; The risk factor set is formed by performing standardized integration based on the factor association table.

3. The method according to claim 1, characterized in that, The step of assessing the critical shift risk contribution of the displacement risk characteristic factors to generate a risk contribution level includes: Extract the offset critical records of the displacement risk characteristic factors; Based on the offset critical record, the tube placement duration attenuation resistance feature is identified to obtain the critical approach value; The critical convergence values ​​are classified into different levels to generate contribution classification results; The risk contribution level is determined based on the contribution grading results.

4. The method according to claim 1, characterized in that, The risk intervention plan generated based on the difference in execution factor weights according to the risk contribution level includes: High-contribution factor items are selected based on the aforementioned risk contribution level; The high-contribution factor items are used to perform a cumulative intervention priority query across nursing cycles to obtain the intervention configuration that can be reused; Based on the aforementioned reusable intervention configuration, a factor weight update table is generated by weight update mapping. Risk intervention plans are generated by matching nursing resource constraints based on the factor weight update table.

5. The method according to claim 1, characterized in that, The process of generating a shift risk profile based on the set of sensitive shift parameters by performing benchmark archiving and storage includes: Individual baseline deviations are compared to extract a subset of the difference parameters from the set of sensitive shift parameters; Based on the subset of differential parameters, differential parameters are categorized to generate differential parameter groups; Historical record tracing is performed on the aforementioned difference parameter groups to generate a parameter tracing set; Based on the parameter traceability set, patient information is associated with risk prediction and labeling to form a displacement risk profile.

6. The method according to claim 1, characterized in that, The step of generating a trend-weighted identifier based on the time-series drift features extracted from the displacement risk association chain includes: Based on the aforementioned displacement risk association chain, extract the time-series scoring records of each scoring node; The drift slope is calculated to obtain the score drift rate from the time-series score records; Based on the rating drift rate, a trend intensity label is generated by classifying the trend intensity according to the cumulative drift. A trend-weighted identifier is generated by performing a weighted coefficient mapping based on the trend strength identifier.

7. The method according to claim 1, characterized in that, The process of defining tiered monitoring ranges and generating tiered early warning identifiers based on the fluctuation range of the shift risk probability value includes: The displacement risk probability value is segmented according to the monitoring time window to generate a continuous time period probability sequence; The fluctuation range distribution is obtained by statistically analyzing the fluctuation amplitude associated with nursing operations on the probability sequence of the continuous time period. Based on the distribution of the fluctuation range, monitoring priorities are allocated to generate a stratified patient list; The stratification warning indicator is determined based on the stratified patient list.

8. The method according to claim 4, characterized in that, The step of querying the cumulative intervention priority across nursing cycles for the high-contribution factor items to obtain the reusable intervention configuration includes: The intervention priority hierarchy structure of the high-contribution factor items is analyzed to generate factor hierarchy relationships; Reference intervention templates are obtained by locating reference intervention templates based on the aforementioned factor hierarchy relationship; The reference intervention template is backtested to verify the occurrence rate of displacement events to obtain valid template records; Based on the effective template records, an adaptability screening is performed to form an intervention configuration that can be reused.

9. The method according to claim 6, characterized in that, The step of calculating the drift slope of the time-series scoring records to obtain the scoring drift rate includes: The time-series scoring records are statistically analyzed to generate interval distribution characteristics by scoring node interval statistics; Based on the interval distribution characteristics, identify score abrupt change drift segments and generate acceleration segment identifiers; Slope fitting analysis is performed on the acceleration section markers to generate slope fitting parameters; The score drift rate is calculated by segmenting the slope breakpoints based on the slope fitting parameters.

10. A catheter tip displacement risk assessment system, characterized in that, include: The risk factor module is used to collect patient position change frequency data and catheter historical positioning deviation records, and to perform clinical risk factor extraction to form a risk factor set based on the rhythm disorder characteristics of the position change frequency data; The risk assessment module is used to extract displacement risk characteristic factors using the risk factor set and the catheter historical positioning offset record, perform critical displacement risk contribution assessment on the displacement risk characteristic factors to generate risk contribution level, and perform factor weight difference mapping based on the risk contribution level to generate risk intervention plan; The file storage module is used to extract scoring indicators from the risk intervention plan to form a shift scoring rule library, use the shift scoring rule library to identify a set of sensitive shift parameters, and perform benchmark archiving and storage based on the set of sensitive shift parameters to generate a shift risk file; The trend arrangement module is used to perform scoring item association parsing on the displacement risk file to extract displacement risk association chains, extract time-series drift features based on the displacement risk association chains to generate trend weighted identifiers, and arrange the trend weighted identifiers according to the risk priority to form a graded displacement assessment sequence. The stratified early warning module is used to perform displacement probability calculation on the stratified displacement assessment sequence to obtain displacement risk probability value, delineate the stratified monitoring range based on the fluctuation range of the displacement risk probability value to generate a stratified early warning identifier, and perform re-examination frequency matching based on the stratified early warning identifier and the risk factor set to generate catheter tip displacement risk assessment result.

Images