A nursing shift handover information association method and system based on a dynamic knowledge graph

Abstract

The application relates to a nursing shift information association method and system based on a dynamic knowledge graph, and belongs to the field of data processing. The method comprises the following steps: integrating patient multi-dimensional data and outputting a situation matrix, wherein the patient multi-dimensional data comprises a patient basic portrait, dynamic indexes, a treatment process and a nursing focus, and the situation matrix comprises acuteness, vulnerability and intervention priority; obtaining and analyzing nursing shift information text, outputting a clinical event set with a corrected potential energy value based on the situation matrix and the nursing shift information text; performing direct effect projection and risk transmission modeling based on the clinical event set, and outputting a time-effect curve and a risk event set carrying a risk value; and constructing a dynamic knowledge graph based on the clinical event set, the time-effect curve and the risk event set. The application realizes the intelligent association of nursing shift information by integrating multi-source clinical data through a dynamic knowledge graph.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data processing, and particularly relates to a nursing shift information association method and system based on a dynamic knowledge graph. BACKGROUND

[0002] The current clinical nursing shift system generally relies on free text recording and oral handover, which has significant defects: information fragmentation leads to missing of key risk points (such as patient night refusal to turn over is only recorded as a behavior event, without being associated with patient medical history and subsequent possible pressure ulcer risk); static data integration is insufficient (vital signs, nursing record text, and risk assessment scale of electronic medical records are stored in scattered places, lacking a dynamic association mechanism); early warning is lagging (such as wound infection relies on naked eye observation of exudation or swelling, missing the golden intervention window). In recent years, knowledge graph and retrieval enhancement generation technology has been gradually applied in the field of clinical nursing shift, but static knowledge graph lacks time sequence dynamic modeling capability and cannot capture the risk evolution trajectory, so there is an urgent need for a nursing shift information association method based on a dynamic knowledge graph which can convert discrete nursing events into quantifiable causal chains. SUMMARY

[0003] To solve the above problems in the prior art, the application provides a nursing shift information association method and system based on a dynamic knowledge graph.

[0004] The object of the application can be achieved by the following technical solutions.

[0005] A nursing shift information association method based on a dynamic knowledge graph, the implementation of the nursing shift information association method comprises the following steps:

[0006] S1: integrating patient multi-dimensional data and outputting a context matrix, the patient multi-dimensional data comprising patient basic profile, dynamic indicators, treatment process and nursing focus, the context matrix comprising acuteness, vulnerability and intervention priority;

[0007] S2: acquiring and analyzing nursing shift information text, and outputting a clinical event set with a corrected potential energy value based on the context matrix and the nursing shift information text;

[0008] S3: performing direct effect projection and risk transmission modeling based on the clinical event set, and outputting a time-effect curve and a risk event set carrying a risk value;

[0009] S4: constructing a dynamic knowledge graph based on the clinical event set, the time-effect curve and the risk event set.

[0010] Preferably, the step S2 specifically comprises:

[0011] parsing the nursing shift information text into the clinical event set, and decomposing the clinical event set into a multi-dimensional structure, the multi-dimensional structure comprising a nursing object, a behavior action, a state transition, a time stamp, and a clinical potential energy change, the clinical potential energy change carrying the revised potential energy value.

[0012] Preferably, the revised potential energy value in the step S2 is specifically obtained as follows:

[0013] obtaining a basic risk coefficient, a behavior intensity coefficient, a context information amount, a time difference, and a department nursing rhythm, and calculating a clinical potential energy index, which is mathematically described as wherein, is the clinical potential energy index, is the basic risk coefficient, is the behavior intensity coefficient, is the context information amount, is the time difference, is the department nursing rhythm.

[0014] obtaining a revised potential energy value based on the clinical potential energy index and the situation matrix, which is mathematically described as wherein, is the revised potential energy value, , and are weights.

[0015] Preferably, the process of the direct effect projection in the step S3 is specifically as follows:

[0016] identifying an operation event set related to medical operations in the clinical event set, and generating a time-effect curve {(t1, effect1), (t2, effect2), (t3, effect3)…} based on the operation event set, wherein t1, t2… are time points, and effect1, effect2… are operation effects corresponding to the time points.

[0017] Preferably, the process of the risk transmission modeling in the step S3 is specifically as follows:

[0018] generating an S-shaped growth function based on a risk growth rate, which is mathematically described as wherein, is the S-shaped growth function, is a risk threshold time, k is the risk growth rate, and t is a current time point.

[0019] obtaining a risk conversion coefficient, and obtaining a risk value based on the risk conversion coefficient and the S-shaped growth function, which is mathematically described as wherein, is the risk value at time t, is the revised potential energy value. is a risk conversion coefficient, the risk event set is obtained based on the risk value.

[0020] Preferably, the construction rule of the dynamic knowledge graph in the step S4 is:

[0021] S401: setting a clinical event as an event node, a risk event as a risk node, and a performed nursing operation as an intervention node, the event node weight of the event node being the correction potential value corresponding to the current time point, the risk node weight of the risk node being the risk value corresponding to the current time point, and the intervention node weight of the intervention node being the operation effect corresponding to the current time point;

[0022] S402: establishing node association according to a preset association type, the association type including time sequence association, causal association, risk transmission, and intervention response;

[0023] S403: starting from any node, traversing all associated paths to form the dynamic knowledge graph and obtaining path weights;

[0024] S404: presetting a path weight threshold, marking a key path based on the path weight and the path weight threshold, and generating a pop-up window prompt.

[0025] Preferably, the path weight is obtained by: obtaining a node distance, the node distance being the number of days between a node timestamp and the current time point, obtaining the path weight according to the node distance and the node weight, and the mathematical description of the path weight being , wherein, is the path weight, w is the node weight, and d n is the node distance, is the path half-life.

[0026] A nursing shift information association system based on a dynamic knowledge graph, used to execute the nursing shift information association method described above, comprising a context matrix output module, a clinical event analysis module, a risk event analysis module, and a graph construction module.

[0027] The context matrix output module is used to integrate patient multi-dimensional data and output a context matrix, the patient multi-dimensional data including patient basic portraits, dynamic indicators, treatment processes, and nursing focuses, and the context matrix including acuteness, vulnerability, and intervention priority.

[0028] The clinical event analysis module is used to acquire and analyze nursing shift information text, and output a clinical event set with a correction potential value based on the context matrix and the nursing shift information text.

[0029] The risk event analysis module is used for direct effect projection and risk conduction modeling based on the clinical event set, and outputs a time-effect curve and a risk event set carrying a risk value;

[0030] The graph construction module is used for constructing a dynamic knowledge graph based on the clinical event set, the time-effect curve and the risk event set.

[0031] The present application has the following beneficial effects:

[0032] (1) Various fragmented information in the nursing shift process is integrated and associated through the dynamic knowledge graph, so that key risk points in the process are avoided.

[0033] (2) The patient's condition is dynamically monitored and predicted through the situation matrix, the correction potential value and the risk value, so as to facilitate medical staff to respond to patient needs in a timely manner.

[0034] (3) Key events are marked through path weight, so as to facilitate medical staff to handle in a timely manner. BRIEF DESCRIPTION OF DRAWINGS

[0035] In order to facilitate those skilled in the art to understand, the present application will be further described below with reference to the accompanying drawings.

[0036] Figure 1 A step flow chart of a nursing shift information association method based on a dynamic knowledge graph. DETAILED DESCRIPTION

[0037] In order to better understand the present application, various aspects of the present application will be described in more detail with reference to the accompanying drawings. It should be understood that these detailed descriptions are only descriptions of exemplary embodiments of the present application, and do not limit the scope of the present application in any way. Throughout the specification, the expression "and / or" includes any and all combinations of one or more of the associated listed items. As used herein, the words "approximately", "about", and similar words are used as approximate terms, not as terms of degree, and are intended to indicate inherent deviations in measured or calculated values that would be recognized by one of ordinary skill in the art. In addition, in the present application, the order of the step processes described does not necessarily represent the order in which the processes appear in actual operation, unless otherwise explicitly limited or derivable from the context.

[0038] It should also be understood that any reference to or discussion of a term or expression in this specification refers to the meaning of that term or expression in the context of the present invention and not in the context of the prior art. It should also be understood that the use of the singular includes the plural unless otherwise expressly stated and that both singular and plural aspects also comprise mixtures thereof. It should also be understood that the use of "or" means "and / or" unless otherwise stated. It should also be understood that the use of "comprise", "comprises", "comprising", "containing", "includes", "including" or "have" means "including but not limited to" unless otherwise stated.

[0039] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

[0040] It should be noted that the embodiments and features of the present application can be combined with each other, if not in conflict. The present application will be described in detail below with reference to the accompanying drawings and embodiments.

[0041] Example 1:

[0042] Please refer to Figure 1 A nursing shift information association method based on a dynamic knowledge graph, comprising:

[0043] S1: integrate patient multi-dimensional data and output a context matrix, the patient multi-dimensional data including patient basic profile, dynamic indicators, treatment progress and nursing focus, the patient basic profile including but not limited to age, diagnosis, allergy history (such as "68 years old, diabetes, penicillin allergy"), the dynamic indicators including but not limited to past 24-hour vital sign trend (such as "blood glucose fluctuation range 4-7 mmol / L"), the treatment progress including but not limited to current treatment plan stage (such as "3rd day after operation / wound healing period"), the nursing focus including but not limited to department focus question (such as "endocrinology department: blood glucose stability priority"), the context matrix including acuteness, vulnerability and intervention priority, the acuteness being (current abnormal indicator number / total indicator number) x severity coefficient x 0.8, the severity coefficient having an open interval (0-1.2) value range, the vulnerability being min[1.5, (age coefficient + complication number x 0.3) x treatment stage coefficient], the intervention priority being nursing focus weight x min(1, recent event density / 6), the mathematical description of the context matrix being ; Example: A 68-year-old diabetic patient with new wound infection, accompanied by complications of kidney disease, abnormal blood glucose, elevated body temperature, and increased white blood cell count. In the past 6 hours, the patient has experienced 3 abnormal events, then the patient's acuity = (3 / 5) x 1.1 x 0.8 = 0.53 (3 of 5 indicators are abnormal, and more serious), vulnerability = (1.2 + 2 x 0.3) x 0.9 = 1.26 (age > 65 age coefficient 1.2, there are two complications of diabetes and kidney disease), intervention priority = 0.8 x min(1, 3 / 6) = 0.4 (blood glucose management weight 0.8, 3 indicator abnormal events in the past 6 hours), and the patient's situation matrix CM = 0.53 x 1.26 x 0.4 = 0.27. .

[0044] S2: Obtain and parse the nursing shift information text, and output a clinical event set with a corrected potential energy value based on the situation matrix and the nursing shift information text;

[0045] S3: Based on the clinical event set, direct effect projection and risk transmission modeling are performed, and a time-effect curve and a risk event set carrying a risk value are output;

[0046] S4: Based on the clinical event set, the time-effect curve and the risk event set, a dynamic knowledge graph is constructed.

[0047] In this embodiment, based on the situation matrix and the nursing shift information text, a clinical event set with a corrected potential energy value is output, which can be implemented by the following steps:

[0048] The nursing shift information text is parsed into the clinical event set by the semantic parser, and the clinical event set is decomposed into a multi-dimensional structure, the multi-dimensional structure including a nursing object, a behavior action, a state transition, a timestamp and a clinical potential energy change, the clinical potential energy change carrying the corrected potential energy value; Example: part of the nursing shift information text is “ordinary ward 3 patients refuse to turn over (10:00), sacrococcygeal skin redness (11:00)”; the nursing shift information text is parsed into a clinical event set = {clinical event A: ordinary ward 3 patients refuse to turn over; clinical event B: sacrococcygeal skin redness}, each clinical event is decomposed into a four-dimensional structure: clinical event A: {nursing object: “ordinary ward 3 patients”, behavior action: “refuse”, state transition: “decrease in activity”, timestamp: “10:00”, clinical potential energy change: “ =0.97”}; clinical event B: {nursing object: “sacrococcygeal”, behavior action: “present”, state transition: “skin from normal to red”, timestamp: “11:00”, clinical potential energy change: “ =0.88”}.

[0049] In the present embodiment, the obtaining of the modified potential energy value is specifically as follows:

[0050] S201: Obtain the basic risk coefficient, the behavior intensity coefficient, the context information amount, the time difference, and the department nursing rhythm, and calculate the clinical potential energy index, which is mathematically described as wherein, is the clinical potential energy index, is the basic risk coefficient (from a medical knowledge base), is the behavior intensity coefficient (the intensity coefficient of the patient behavior action), is the context information amount (i.e., the number of associated symptoms), is the time difference (in hours, i.e., the time difference from the last related event), is the department nursing rhythm (in hours, i.e., the time interval for routine nursing / patrol in each department);

[0051] S202: Obtain the modified potential energy value based on the clinical potential energy index and the situation matrix, which is mathematically described as wherein, is the modified potential energy value, , and are weights.

[0052] In the present embodiment, the purpose of the direct effect projection is to quantify the immediate treatment effect of the nursing operation (such as the drug onset process), and the process is specifically as follows:

[0053] Identify the operation event set related to the medical operation in the clinical event set, and generate the time-effect curve {(t1, effect1), (t2, effect2), (t3, effect3)…} based on the operation event set, wherein t1, t2… are time points, and effect1, effect2… are operation effects at the corresponding time points, which are mathematically described as wherein, is the operation intensity coefficient (obtained from the drug dose, for example, 20 mg furosemide is 1.0, 10 mg furosemide is 0.5), is the effect function at time t (for example, rapid-acting insulin =e^(-0.15t)), is the maximum effect reference value (for example, hypoglycemic drug =10 (blood glucose decrease of 10 mmol / L / h)). If there is no medical operation, output is empty.

[0054] In the present embodiment, the purpose of the risk transmission modeling is to predict the dynamic evolution trajectory of the potential nursing risk (such as the infection deterioration trend), and the process is specifically as follows:

[0055] Generate an S-shaped growth function based on the risk growth rate, mathematically described as ,in, is an S-shaped growth function, is the risk threshold time (the time point when the risk begins to accelerate, for example: wound infection = 24h, which is the starting point of the logarithmic phase of bacterial proliferation), k is the risk growth rate (obtained by disease severity classification, e.g., diabetic foot ulcer: mild k = 0.3, severe k = 0.6);

[0056] The risk conversion coefficient is obtained by analyzing the hospital adverse event database (for example, if 100 cases of refusal to turn over are analyzed and the proportion of pressure sores occurring within 72 hours is 70%, then the risk conversion coefficient is 0.7). The risk value is obtained based on the risk conversion coefficient and the S-type growth function, which is mathematically described as follows: ,in, is the risk value at time t, is the risk conversion coefficient, It is an S-shaped growth function. Based on the risk value, the risk event set is obtained. The mathematical description is risk event set = {risk event a, ,Timestamp; risk event b, ,timestamp;…}( is the risk value set of risk event a in the entire time process, is the risk value set of risk event b in the entire time process); the risk events here are events that may be caused by nursing operations / patient behaviors, not events that have already occurred or will definitely occur. For example, a patient's refusal to turn over may lead to a pressure sore at some point in the future.

[0057] Example: Event A: {Nursing object: "Patient in bed 3 of general ward", behavioral action: "Refuse debridement", state transition: "Increased wound exudate", clinical potential change: =0.89", its direct effect projection is: no nursing operation, output is empty; its risk conduction model is: the risk conversion coefficient is 0.75, at this time, the risk value needs to be predicted after 24 hours and 36 hours respectively, and the risk growth rate k=0.5, then , .

[0058] In this embodiment, the construction rules of the dynamic knowledge graph in step S4 are:

[0059] S401: Set the clinical event as the event node, the risk event as the risk node, and the executed nursing operation as the intervention node. The event node weight of the event node is the modified potential energy value corresponding to the current time point, the risk node weight of the risk node is the risk value corresponding to the current time point, and the intervention node weight of the intervention node is the operation effect corresponding to the current time point (i.e., the corresponding position of the time-effect curve);

[0060] S402: Establish node associations based on preset association types, including temporal association (event A occurs before event B), causal association (event A directly causes event B), risk transmission (clinical event A triggers risk event B), and intervention response (nursing operation A targets risk event B). The establishment of node associations is specifically as follows: (1) Temporal association is specifically applicable to scenarios where events occur in chronological order but the causal relationship is unclear, such as: Node A: 10:00 The patient refuses to turn over, Node B: 14:00 The patient's skin becomes red, but it is not clear that the redness is caused by refusing to turn over, then nodes AB are temporally associated with the time nodes. (2) Causal association is specifically applicable to events with clear clinical causal relationships. The established rules are to match the clinical causal knowledge base (such as "the probability of hypoglycemia caused by not eating on time is 62%); meet the time sequence (cause first, effect later); and conform to the physiological and pathological mechanism. For example: Node A: At 8:00, the patient refused to take hypoglycemic drugs, Node B: At 12:00, the patient's blood sugar increased significantly. The clinical causal knowledge base query shows that the probability of a causal relationship is 62%, so nodes AB are causally associated. (3) Risk conduction is specifically applicable to scenarios where current events trigger future risks. The established rule is that node A is an actual event; node B is a risk event predicted by S3. For example: Node A: At 9:00, the patient refused wound care, Node B: The 24h infection risk is predicted to increase, so nodes AB are associated with risk conduction. (4) Intervention response is specifically applicable to scenarios where nursing measures target specific risks. The established rule is: Node A is a nursing intervention measure; Node B is a risk event in S3. For example: Node A: Debridement is performed at 11:00, Node B: The risk of wound infection within 24 hours is 0.45, so nodes AB are associated with intervention response;

[0061] S403: Starting from any node, traverse all associated paths to form the dynamic knowledge graph and obtain path weights;

[0062] The path weight is obtained by obtaining the node distance, which is the distance in days between the node timestamp and the current time point, and obtaining the path weight according to the node distance and the node weight. The node weight is the event node weight, risk node weight and intervention node weight. The mathematical description of the path weight is ,in, is the path weight, w is the node weight, dn For node distance, For path half-life (acute problem =2 days, chronic problem =7 days). At this time, the timeline can be dragged to view the changes in path weight and node weight.

[0063] S404: preset path weight threshold, mark key path based on the path weight and the path weight threshold, and generate a pop-up prompt (example: Greater than 1.5 is highlighted and immediately pushed to the nurse's mobile phone).

[0064] Embodiment 2:

[0065] A nursing shift information association system based on a dynamic knowledge graph includes a context matrix output module, a clinical event analysis module, a risk event analysis module, and a graph construction module.

[0066] The context matrix output module is used to integrate patient multi-dimensional data and output a context matrix, the patient multi-dimensional data including patient basic profile, dynamic indicators, treatment progress, and nursing focus, and the context matrix including acuteness, vulnerability, and intervention priority.

[0067] The clinical event analysis module is used to acquire and analyze nursing shift information text, and output a clinical event set with a corrected potential energy value based on the context matrix and the nursing shift information text.

[0068] The risk event analysis module is used to perform direct effect projection and risk transmission modeling based on the clinical event set, and output a time-effect curve and a risk event set carrying a risk value.

[0069] The graph construction module is used to construct a dynamic knowledge graph based on the clinical event set, the time-effect curve, and the risk event set.

[0070] The above is only a preferred embodiment of the present application, and is not intended to limit the present application in any form. Although the present application has been disclosed as above with a preferred embodiment, it is not intended to limit the present application. Any person skilled in the art can make some changes or modifications to the above disclosed technical content to make equivalent embodiments with equivalent changes, without departing from the scope of the technical solution of the present application. Any simple modification, equivalent change, and modification of the above embodiments based on the technical essence of the present application are still within the scope of the technical solution of the present application.

Claims

1. A method for associating nursing handover information based on dynamic knowledge graph, characterized in that: The implementation of the nursing handover information association method includes the following steps: Step S1: Integrate multidimensional patient data and output a context matrix, wherein the multidimensional patient data includes a basic patient profile, dynamic indicators, treatment progress, and nursing focus, and the context matrix includes acuity, vulnerability, and intervention priority; Step S2: Acquire and parse the nursing handover information text, and output a clinical event set with a modified potential energy value based on the situation matrix and the nursing handover information text; Step S3: Perform direct effect projection and risk conduction modeling based on the clinical event set, and output a time effect curve and a risk event set carrying risk values. The specific process of direct effect projection is as follows: Identify an operational event set related to medical operations in the clinical event set, and generate a time-effect curve {(t1, effect1), (t2, effect2), (t3, effect3)...} based on the operational event set, where t1, t2... are time points, and effect1, effect2... are operational effects at corresponding time points; Step S4: constructing a dynamic knowledge graph based on the clinical event set, the time effect curve and the risk event set. The construction rules of the dynamic knowledge graph are: S401: Set the clinical event as the event node, the risk event as the risk node, and the executed nursing operation as the intervention node. The event node weight of the event node is the modified potential energy value corresponding to the current time point, the risk node weight of the risk node is the risk value corresponding to the current time point, and the intervention node weight of the intervention node is the operation effect corresponding to the current time point. S402: Establishing node associations according to preset association types, wherein the association types include temporal association, causal association, risk transmission, and intervention response; S403: Starting from any node, traverse all associated paths to form the dynamic knowledge graph and obtain path weights; S404: Preset a path weight threshold, mark key paths based on the path weight and the path weight threshold, and generate a pop-up prompt.

2. The nursing handover information association method according to claim 1, characterized in that: The step S2 is specifically as follows: The nursing handover information text is parsed into the clinical event set, and the clinical event set is decomposed into a multidimensional structure, wherein the multidimensional structure includes nursing objects, behavioral actions, state transitions, timestamps, and clinical potential energy changes, and the clinical potential energy changes carry the modified potential energy value.

3. The nursing handover information association method according to claim 2, characterized in that: The acquisition of the modified potential energy value in step S2 is specifically as follows: Obtain the basic risk coefficient, behavior intensity coefficient, context information, time difference and department nursing rhythm, and calculate the clinical potential index, which is mathematically described as ,in, is the clinical potential index, is the basic risk factor, is the behavior intensity coefficient, is the amount of context information, is the time difference, Provides the rhythm of nursing care for the department; Based on the clinical potential index and the situation matrix, a modified potential value is obtained, which is mathematically described as ,in, To correct the potential energy value, 、 and is the weight.

4. The nursing handover information association method according to claim 1, characterized in that: The specific process of risk transmission modeling in step S3 is as follows: Generate an S-shaped growth function based on the risk growth rate, mathematically described as ,in, is an S-shaped growth function, is the risk threshold time, k is the risk growth rate, and t is the current time point; Obtain the risk conversion coefficient, and obtain the risk value based on the risk conversion coefficient and the S-type growth function, which is mathematically described as ,in, is the risk value at time t, To correct the potential energy value, is the risk conversion coefficient, and the risk event set is obtained based on the risk value.

5. The nursing handover information association method according to claim 1, characterized in that: The path weight is obtained by obtaining the node distance, which is the distance in days between the node timestamp and the current time point, and obtaining the path weight according to the node distance and the node weight. The mathematical description of the path weight is: ,in, is the path weight, w is the node weight, d n is the node spacing, is the path half-life.

6. A nursing handover information association system based on dynamic knowledge graph, characterized in that: The system is used to execute the nursing handover information association method according to any one of claims 1 to 5, comprising a situation matrix output module, a clinical event analysis module, a risk event analysis module and a graph construction module; The situation matrix output module is used to integrate multi-dimensional patient data and output a situation matrix, wherein the multi-dimensional patient data includes a basic patient profile, dynamic indicators, treatment progress, and nursing focus, and the situation matrix includes acuity, vulnerability, and intervention priority; The clinical event parsing module is used to obtain and parse the nursing handover information text, and output a clinical event set with a corrected potential energy value based on the situation matrix and the nursing handover information text; The risk event analysis module is used to perform direct effect projection and risk conduction modeling based on the clinical event set, and output a time effect curve and a risk event set carrying a risk value; The graph construction module is used to construct a dynamic knowledge graph based on the clinical event set, the time effect curve and the risk event set.

Images