A postoperative complication risk prediction method for surgical care

Abstract

The present application relates to the technical field of complication risk assessment, in particular to a postoperative complication risk prediction method for surgical nursing. The present application firstly analyzes the correlation coefficient between indexes and complications based on a large amount of postoperative nursing information of historical patients, and constructs a causal network of complications; then analyzes the change correlation between the monitoring information of the indexes corresponding to the adjacent nodes on each causal chain, and the correlation coefficient between the indexes corresponding to the nodes and the complications, obtains the connection coefficient between the adjacent nodes and updates the causal network; finally, the risk of each postoperative complication is predicted in combination with the current postoperative nursing information of the current patient. The present application constructs a risk-driven causal network accurately reflecting different complications based on a large amount of postoperative nursing reference provided by a large number of historical patients, so that the postoperative complication risk of the patient can be accurately predicted by means of the risk transmission causal chain of the complications in the causal network, and the complication risk prediction effect is improved.

Description

Technical Field

[0001] This invention relates to the field of complication risk assessment technology, and more specifically to a method for predicting the risk of postoperative complications in surgical care. Background Technology

[0002] In surgical care, postoperative complications (such as infection, bleeding, deep vein thrombosis, etc.) are key factors affecting the patient's recovery process and increasing the medical burden. They are closely related to the patient's underlying diseases, intraoperative procedures, and postoperative care. Accurately predicting the risk of complications and intervening in advance can significantly reduce the incidence and shorten the length of hospital stay, which is of great significance for improving the quality of surgical care.

[0003] Most existing methods for predicting postoperative complication risks are limited to simply judging whether a risk will occur. They cannot delve into the interactions and transmission pathways of different factors in the risk formation process, nor can they reflect the dynamic characteristics of risk changes over time and with various intervention factors. As a result, the prediction results remain at the level of risk warning and are difficult to correlate with actual nursing procedures, thus affecting the accuracy and effectiveness of postoperative complication risk prediction. Summary of the Invention

[0004] To address the technical problem of ineffective prediction of postoperative complication risks, the present invention aims to provide a method for predicting postoperative complication risks in surgical care, the specific technical solution of which is as follows: Acquire monitoring information for each indicator during postoperative care for each historical patient, as well as the timing of the occurrence of each complication; Based on the temporal distribution of the occurrence time of each complication in all historical patients, and the distribution of abnormal deviations in the temporal data of each indicator, the correlation coefficient between each indicator and each complication is obtained. A causal network of complications is constructed based on the monitoring information of all historical patients' corresponding indicators and complications. The last node of each causal chain in the causal network is a complication, and each non-last node corresponds to at least one indicator. The connection coefficient between adjacent nodes is obtained and the causal network is updated based on the change correlation between the monitoring information of the indicators corresponding to adjacent nodes in each causal chain and the correlation coefficient between the indicator corresponding to each node and the complication corresponding to the last node. Based on the current monitoring information of each indicator of the current patient during postoperative care, and the connection coefficients between adjacent nodes in the updated causal network, the risk prediction probability of each postoperative complication of the current patient is obtained.

[0005] Furthermore, the method for obtaining the correlation coefficient includes: The postoperative care process is divided into monitoring sub-periods of preset duration; For each complication, the frequency of occurrence of the complication within each monitoring sub-period is determined based on the time of occurrence of the complication in all historical patients, and a distribution sequence of the occurrence of the complication is constructed based on the frequency of occurrence within each monitoring sub-period. For each indicator, in each monitoring sub-period, based on the deviation of the concentrated characteristics of the monitoring information of each historical patient relative to the corresponding preset standard, the abnormal deviation parameter of the indicator is obtained, and the abnormal distribution sequence of the indicator is constructed based on the abnormal deviation parameter of the indicators of all historical patients in each monitoring sub-period. Based on the correlation between the abnormal distribution sequence of each indicator and the occurrence distribution sequence of each complication, the correlation coefficient between each indicator and each complication is obtained.

[0006] Furthermore, the method for obtaining the abnormal distribution sequence includes: For each indicator, in each monitoring sub-period, historical patients whose abnormal deviation parameter is greater than the corresponding preset threshold are regarded as abnormal patients, and the frequency of occurrence of abnormal patients in each monitoring sub-period is used as sequence elements to construct an abnormal distribution sequence.

[0007] Furthermore, the method for obtaining the connection coefficients includes: On each causal chain, the connection rigidity coefficient between adjacent nodes is obtained based on the change correlation between the monitoring information of the corresponding indicators of adjacent nodes and the difference in the number of corresponding indicators of adjacent nodes. In each causal chain, the average of the correlation coefficients between each indicator corresponding to each node and the complication corresponding to the last node is used as the node weight of each node; the connection weight between adjacent nodes is obtained based on the difference in the node weights of adjacent nodes. The connection rigidity coefficient is weighted using the connection weights, and the normalized value of the weighted result is used as the connection coefficient between the corresponding adjacent nodes.

[0008] Furthermore, the method for obtaining the connection stiffness coefficient includes: For each adjacent node, the indicator corresponding to the preceding node is taken as the target indicator, and the indicator corresponding to the following node is taken as the reference indicator. Based on the set characteristics of the Pearson correlation coefficient between the abnormal distribution sequences corresponding to each target indicator and each reference indicator, the rigid parameters between adjacent nodes are obtained. The ratio of the number of target indicators to the number of reference indicators is taken as the rigid weight between adjacent nodes, where the number of target indicators is the numerator. The rigid parameters are weighted using the rigid weight, and the weighted result is taken as the connection rigidity coefficient between the corresponding adjacent nodes.

[0009] Furthermore, the method for obtaining the connection weights includes: In adjacent nodes, the ratio of the node weight of the preceding node to the node weight of the following node is used as the connection weight between adjacent nodes, where the node weight of the preceding node is the denominator.

[0010] Furthermore, methods for updating causal networks include: In a causal network, edges between adjacent nodes whose connection coefficients are less than a preset threshold are deleted, and nodes with no connection are removed to obtain an updated causal network.

[0011] Furthermore, the method for obtaining the risk prediction probability includes: Based on the current monitoring information of each indicator during the postoperative care of the current patient, all reference causal chains of the current patient are determined in the updated causal network; the risk prediction probability of the corresponding complication is obtained according to the connection coefficient between adjacent nodes in each reference causal chain.

[0012] Furthermore, the risk prediction probability of the corresponding complication is obtained based on the connection coefficients between adjacent nodes in each reference causal chain, including: The cumulative product of the connection coefficients between all adjacent nodes in each reference causal chain is normalized to obtain the risk prediction probability of the corresponding complication.

[0013] Furthermore, the indicators include at least each vital sign and each nursing procedure.

[0014] The present invention has the following beneficial effects: This invention first acquires monitoring information for each indicator and the occurrence time of each complication during the postoperative care of each historical patient, preparing for the construction of a causal network for subsequent analysis. Then, based on the temporal distribution of the occurrence time of each complication across all historical patients, and the distribution of abnormal deviations in the temporal sequence of monitoring information for each indicator, the impact of abnormal deviations on the occurrence of complications is analyzed, and the correlation coefficient between each indicator and each complication is obtained. Next, a causal network of complications is constructed based on the monitoring information of corresponding indicators and complications for all historical patients. The influence relationship between nodes is analyzed based on the changes in monitoring information of indicators corresponding to adjacent nodes in each causal chain. Furthermore, the progressive influence relationship of adjacent nodes on the development of complications is analyzed by combining the correlation coefficient between the indicator corresponding to each node and the complication corresponding to the last node, thereby comprehensively obtaining the connection coefficient between adjacent nodes and updating the causal network and optimizing its structure. Finally, based on the current monitoring information of each indicator during the postoperative care of the current patient, and the connection coefficient between adjacent nodes in the updated causal network, the risk prediction probability of each postoperative complication for the current patient is obtained. Based on a large number of historical patients' postoperative care references, this invention constructs a risk-driven causal network that accurately reflects different complications. This allows for the accurate prediction of patients' postoperative complication risks by leveraging the risk transmission causal chain of complications within the causal network, thereby improving the effectiveness of risk prediction. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating a method for predicting the risk of postoperative complications in surgical care, provided as an embodiment of the present invention; Figure 2 A flowchart illustrating a method for obtaining correlation coefficients according to an embodiment of the present invention; Figure 3 This is a flowchart illustrating a method for obtaining connection coefficients according to an embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a postoperative complication risk prediction method for surgical care proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following description, in conjunction with the accompanying drawings, details a specific scheme for a method of predicting the risk of postoperative complications in surgical care provided by the present invention.

[0020] Please see Figure 1 The diagram illustrates a flowchart of a method for predicting the risk of postoperative complications in surgical care, provided by an embodiment of the present invention, specifically including: Step S1: Obtain monitoring information for each indicator and the time of occurrence of each complication during the postoperative care process for each historical patient.

[0021] In one embodiment of the present invention, existing Internet of Things devices and nursing record systems are first used to collect monitoring information of each indicator and the occurrence time of each complication during the postoperative care process of each historical patient in real time, so as to provide a reference basis for the subsequent construction of a causal network of complications to analyze and predict the risk of various complications. Considering that complications are mainly affected by the patient's physical condition and postoperative care, good physical condition and reasonable postoperative care will reduce the risk of complications. By monitoring the patient's vital signs and nursing procedures during postoperative care, it is possible to help assess their relationship with complications. Based on this, in a preferred embodiment of the present invention, the indicators include at least each vital sign and each nursing procedure.

[0022] Specifically, vital signs should include at least the following indicators that reflect the patient's physical condition, immune function, and (surgical) incision healing: body temperature, heart rate, blood oxygen, blood pressure, blood glucose, incision effusion, venous blood flow velocity, white blood cell count, and C-reactive protein level. Nursing procedures should include at least the following: changes in incision dressings, administration of postoperative medications such as antibiotics, and the patient's ability to get out of bed. All parameters of each indicator during postoperative care are used as monitoring information. The monitoring information is time-series data, which can help analyze the changes of indicators during postoperative care. The occurrence time of complications is the first discovery time during postoperative care. The types of complications include incision infection, thrombosis, etc. Since the postoperative care duration may vary for different patients, a fixed monitoring period can be defined, such as one week or three days postoperatively. The acquisition of monitoring information for the above indicators is already a current technology, and the specific process will not be elaborated further. Implementers may also add relevant indicators for monitoring.

[0023] It should be noted that the monitoring information under all indicators needs to be cleaned and standardized to remove the dimensions of the indicator parameters under each indicator for subsequent analysis and calculation; data cleaning and standardization are well-known techniques and will not be elaborated further.

[0024] Step S2: Based on the temporal distribution of the occurrence time of each complication in all historical patients, and the distribution of abnormal deviations in the temporal sequence of monitoring information for each indicator, obtain the correlation coefficient between each indicator and each complication.

[0025] After obtaining the monitoring information of each historical patient under each indicator and the time of occurrence of complications, we can further use the references provided by a large number of historical patients to assess the relationship between each indicator and each complication, that is, the possibility that abnormal changes in each indicator will cause each complication, so as to construct the causal network of complications in the future. Furthermore, considering that if there is an influence relationship between the indicator and the complication, the time distribution of abnormal deviations in the monitoring information of all historical patients under the indicator should have a certain similarity to the time distribution of the occurrence of the complication; therefore, this embodiment of the invention will obtain the correlation coefficient between each indicator and each complication based on the temporal distribution of the occurrence time of each complication in all historical patients and the temporal distribution of abnormal deviations in the monitoring information of each indicator. The correlation coefficient reflects the possibility that abnormal changes in the indicator will cause complications.

[0026] Preferably, in one embodiment of the present invention, the method for obtaining the correlation coefficient includes: Please see Figure 2 The flowchart illustrates a method for obtaining correlation coefficients according to an embodiment of the present invention, specifically including: Step S201: Divide the postoperative care process into monitoring sub-periods of preset duration.

[0027] It should be noted that the monitoring information corresponding to the above indicators is time-series data, but the monitoring frequency is limited by factors such as the equipment parameters of the relevant instruments, medical orders or medical resources, which may result in different numbers of indicator parameters in the monitoring information of different indicators.

[0028] In one embodiment of the present invention, the postoperative care process is first divided into several monitoring sub-periods of equal length. Specifically, the preset duration is set to 6 hours, which can also be adjusted by the implementer. This is to prepare for the subsequent statistical analysis of the concentrated characteristics of the monitoring information corresponding to each indicator in each monitoring sub-period, in order to adjust and align the monitoring information corresponding to each indicator, and to avoid the inability to analyze the monitoring information due to the different lengths of each monitoring information.

[0029] Step S202: For each complication, based on the occurrence time of the complication in all historical patients, determine the frequency of the complication in each monitoring sub-period, and construct the complication occurrence distribution sequence based on the frequency of occurrence in each monitoring sub-period.

[0030] Specifically, taking any complication as an example, without going into detail, the occurrence time of all historical patients' complications is mapped to the timeline of the postoperative care process. The number of times the complication occurs in each monitoring sub-period is counted, i.e., the frequency of occurrence. The frequency of occurrence in each monitoring sub-period is used as a sequence element to construct the complication occurrence distribution sequence. The occurrence distribution sequence is based on a large amount of historical patient complication information, reflecting the distribution of complications during postoperative care. This prepares the basis for subsequent assessment of the correlation coefficient between indicators and complications by combining changes in monitoring information corresponding to each indicator.

[0031] Step S203: For each indicator, in each monitoring sub-period, based on the deviation of the concentrated characteristics of the monitoring information of each historical patient relative to the corresponding preset standard, obtain the abnormal deviation parameter of the indicator, and construct the abnormal distribution sequence of the indicator based on the abnormal deviation parameters of the indicators of all historical patients in each monitoring sub-period.

[0032] Specifically, for each indicator, within each monitoring sub-period, the mean is used to represent the central characteristics. First, the mean of the monitoring information (indicator parameters) of each historical patient under this indicator is obtained. Then, the absolute value of the difference is used to measure the deviation between the mean and the preset standard corresponding to the indicator, and the abnormal deviation parameter of the indicator is obtained. For example, regarding body temperature, the absolute value of the difference between all body temperature monitoring values ​​and the preoperative baseline body temperature within a certain monitoring sub-period can be calculated to obtain the abnormal deviation parameter of body temperature; regarding the replacement of incision dressings, the absolute value of the difference between the actual number of times the incision dressing was replaced and the number of times it should have been replaced within a certain monitoring sub-period can be calculated to obtain the abnormal deviation parameter of incision dressing replacement.

[0033] After obtaining the abnormal deviation parameters of each indicator for each historical patient in each monitoring sub-period, we can further evaluate the abnormal deviation distribution of the indicator for all historical patients in each monitoring sub-period, construct an abnormal distribution sequence, and prepare for subsequent analysis of the correlation between the abnormal deviation distribution of the indicator and the occurrence distribution of complications. In a preferred embodiment of the present invention, the method for obtaining the abnormal distribution sequence includes: For each indicator, in each monitoring sub-period, historical patients whose abnormal deviation parameters are greater than the corresponding preset threshold are regarded as abnormal patients, and the frequency of occurrence of abnormal patients in each monitoring sub-period is used as sequence elements to construct an abnormal distribution sequence.

[0034] Specifically, since all monitoring information has been standardized, the preset threshold can be set to 0.1, and the implementer can also adjust the preset threshold according to the actual situation. Then, for each indicator, in each monitoring sub-period, abnormal patients are screened from all historical patients based on the abnormal deviation parameter, and the number of abnormal patients, i.e., the frequency of occurrence, is counted. In the postoperative care process, the frequency of occurrence of abnormal patients in each monitoring sub-period is used as a sequence element to construct an abnormal distribution sequence. The abnormal distribution sequence reflects the abnormal deviations of the parameters of each indicator in a large number of historical patients during the postoperative care process, which prepares for subsequent assessment of the correlation coefficient between the indicators and complications by combining the changes of the corresponding monitoring information of each indicator.

[0035] Step S204: Based on the correlation between the abnormal distribution sequence of each indicator and the occurrence distribution sequence of each complication, obtain the correlation coefficient between each indicator and each complication.

[0036] Specifically, the mean square error between the abnormal distribution sequence of each indicator and the occurrence distribution sequence of each complication is first obtained. The smaller the mean square error, the greater the correlation between the changes in the sequences. The mean square error is then mapped to the exponential function exp(-x) with the natural constant e as the base. The logical relationship and value range are adjusted to obtain the correlation coefficient between each indicator and each complication.

[0037] It should be noted that mean squared error is already existing technology and will not be elaborated further.

[0038] Step S3: Construct a causal network of complications based on the monitoring information of all historical patients' corresponding indicators and complications. The last node of each causal chain in the causal network is a complication, and each non-last node corresponds to at least one indicator. Based on the change correlation between the monitoring information of the indicators corresponding to adjacent nodes in each causal chain, and the correlation coefficient between the indicator corresponding to each node and the complication corresponding to the last node, obtain the connection coefficient between adjacent nodes and update the causal network.

[0039] Considering that the risk-driving pathways for different complications may be different, for example, untimely changes of incision dressings may lead to increased incision exudate, which may in turn lead to bacterial growth and infection; insufficient time for patients to get out of bed after surgery may lead to a decrease in venous blood flow velocity, which may in turn lead to thrombosis; Based on the above examples, it can be inferred that the path of complications is usually "risk exposure - process state - consequence event". By constructing a targeted causal relationship network, the risk source of complications can be identified. Therefore, this embodiment of the invention will construct a causal network of complications based on the monitoring information of all historical patients' corresponding indicators and complications, in order to prepare for subsequent complication risk prediction.

[0040] It should be noted that the construction of causal networks for complications is already a known technology; the general construction process is briefly described here: First, trace the causal chain for each complication, for example: ① Nursing procedures → incision manifestations → infection (e.g., delayed dressing changes → incision exudation → infection); ② Activity status → blood flow → thrombus (e.g., insufficient movement after getting out of bed → decreased venous blood flow velocity → thrombus); ③ Physiological abnormalities → immunity → infection (e.g., postoperative hypothermia → decreased immune function → infection); After sorting out all causal chains, all indicators are manually classified to a certain node in the causal chain. For example, white blood cell count and C-reactive protein level are classified to the node corresponding to "decreased immune function", and so on, to determine all indicators corresponding to each node. Then, common nodes in all causal chains are identified, and paths are added between nodes to obtain a directed acyclic graph and a causal network of complications. Each causal chain in the causal network contains at least 3 nodes, where the last node is a complication. Nodes are abstract states and may require indicators from multiple dimensions to be jointly represented. That is, each non-last node in each causal chain corresponds to at least one indicator.

[0041] Since the causal network is initially constructed from the path of complication, it is impossible to quantify the connection relationship between each node, and thus impossible to infer the risk of each causal chain, making it difficult to predict the risk of complications or to carry out nursing interventions. Based on this, the embodiments of the present invention will further analyze and obtain the connection coefficients between adjacent nodes with connection relationships in the causal network, adaptively update the causal network, and simplify the network structure to improve analysis efficiency.

[0042] Considering that in the causal network of complications, each node in each causal chain corresponds to at least one indicator, by analyzing the correlation between changes in the monitoring information of two indicators corresponding to adjacent nodes, the influence relationship between nodes can be assessed. At the same time, each causal chain corresponds to a complication, and there is a certain correlation between changes between each indicator and each complication. By analyzing the correlation coefficient between each node in the adjacent nodes and the complication corresponding to the last node, it can be further helpful to assess the progressive influence of adjacent nodes on the occurrence of complications. Based on this, the embodiments of the present invention will obtain the connection coefficient between adjacent nodes and update the causal network according to the change correlation between the monitoring information of the corresponding indicators of adjacent nodes in each causal chain and the correlation coefficient between the corresponding indicator of each node and the complication corresponding to the end node; the connection coefficient reflects the progressive influence relationship of adjacent nodes in the causal chain on the occurrence of complications.

[0043] Preferably, in one embodiment of the present invention, the method for obtaining the connectivity coefficients includes: Please see Figure 3 The flowchart illustrates a method for obtaining connectivity coefficients according to an embodiment of the present invention, specifically including: Step S301: On each causal chain, based on the change correlation between the monitoring information of the corresponding indicators of adjacent nodes and the difference in the number of corresponding indicators of adjacent nodes, obtain the connection rigidity coefficient between adjacent nodes.

[0044] Considering that the more the monitoring information of different indicators corresponding to adjacent nodes in the causal chain has a certain correlation of change, that is, the more likely the change of the indicator of the preceding node is to cause the change of the indicator of the following node, it indicates that there is a certain influence relationship between adjacent nodes. Furthermore, considering that the connection relationship between adjacent nodes is directional, the difference in the number of indicators corresponding to the preceding node and the following node can characterize the direction of node state transmission, thereby helping to assess the progressive influence relationship between adjacent nodes from the side. Based on the above logic, the connection rigidity coefficient between adjacent nodes can be obtained by considering the changes in the monitoring information of the corresponding indicators of adjacent nodes in each causal chain, as well as the differences in the number of corresponding indicators of adjacent nodes. The connection rigidity coefficient initially reflects the connection influence between adjacent nodes, preparing for the subsequent determination of the connection coefficient.

[0045] In a preferred embodiment of the present invention, considering that the abnormal distribution sequence of each indicator constructed in step S2 reflects the abnormal changes in indicator parameters to a certain extent, and that the correlation of abnormal changes in indicator parameters of adjacent nodes can provide a reference for node correlation, while the Pearson correlation coefficient can be used to measure the correlation between sequences; furthermore, considering that between adjacent nodes, the more indicators the preceding node has and the fewer indicators the following node has, indicating that the direction of node state transmission is concentrated rather than divergent, it suggests that there is a certain progressive influence relationship between adjacent nodes; based on this, the method for obtaining the connection rigidity coefficient includes: For each adjacent node, the indicator corresponding to the preceding node is taken as the target indicator, and the indicator corresponding to the following node is taken as the reference indicator. Based on the ensemble characteristics of the Pearson correlation coefficient between the abnormal distribution sequences corresponding to each target indicator and each reference indicator, the rigid parameters between adjacent nodes are obtained. The ratio of the number of target indicators to the number of reference indicators is taken as the rigid weight between adjacent nodes, where the number of target indicators is the numerator. The rigid parameters are weighted using the rigid weight, and the weighted result is taken as the connection rigidity coefficient between the corresponding adjacent nodes.

[0046] Specifically, in each pair of adjacent nodes in each causal chain, the indicator corresponding to the preceding node is first used as the target indicator, and the indicator corresponding to the following node is used as the reference indicator; then the target indicators and reference indicators between adjacent nodes are grouped in pairs to calculate the Pearson correlation coefficient between the abnormal distribution sequences corresponding to each target indicator and each reference indicator. Then, the mean is used to represent the central characteristics. The Pearson correlation coefficients between the target indicators and the reference indicators in all groups of adjacent nodes are averaged to obtain the rigid parameters between adjacent nodes. Then, the number of target indicators is used as the numerator, the number of reference indicators is used as the denominator, and the ratio of the fractions is used as the rigid weight between the corresponding adjacent nodes. The rigid parameters are multiplied by the rigid weights to obtain the connection rigidity coefficient between the corresponding adjacent nodes.

[0047] It should be noted that the Pearson correlation coefficient is already existing technology and will not be discussed further.

[0048] Step S302: On each causal chain, the average value of the correlation coefficient between each indicator corresponding to each node and the complication corresponding to the end node is used as the node weight of each node; based on the difference in the node weights of adjacent nodes, the connection weight between adjacent nodes is obtained.

[0049] Considering that in the causal network of complications, each node in each causal chain corresponds to at least one indicator, and each causal chain corresponds to a complication, by analyzing the correlation coefficient between each node in each pair of adjacent nodes and the complication corresponding to the last node, we can help analyze the influence weight of each node on the complication, and thus help to indirectly assess the progressive influence of adjacent nodes on the occurrence of complications and determine the connection weight between adjacent nodes; the connection weight evaluates the connection influence relationship between adjacent nodes from the perspective of the influence of nodes on complications.

[0050] Specifically, in each causal chain, the average correlation coefficient between each indicator corresponding to each node and the complication corresponding to the last node is used as the node weight of each node; the node weight reflects the impact of each node on the complication. Considering that in each pair of adjacent nodes, if the influence of the later node on the complication is greater than that of the earlier node, then the node closer to the onset of the complication has a greater direct influence, and the connection influence of adjacent nodes has greater reference value; therefore, in a preferred embodiment of the present invention, the method for obtaining the connection weight includes: In adjacent nodes, the ratio of the node weight of the preceding node to the node weight of the following node is used as the connection weight between adjacent nodes, where the node weight of the preceding node is the denominator.

[0051] Step S303: The connection rigidity coefficient is weighted using connection weights, and the normalized value of the weighted result is used as the connection coefficient between the corresponding adjacent nodes.

[0052] Since both connection weights and connection rigidity coefficients reflect the connection influence between corresponding adjacent nodes from different perspectives, the connection weights and connection rigidity coefficients are multiplied together, and the product is linearly normalized to obtain the connection coefficients between adjacent nodes.

[0053] It should be noted that linear normalization is performed on the dimension of all adjacent nodes in the causal network, which is an existing technology and will not be elaborated further.

[0054] Considering that there may be weak connections between nodes, the causal network can be updated based on the connection coefficients to optimize the network structure.

[0055] Preferably, in one embodiment of the present invention, the method for updating the causal network includes: In a causal network, edges between adjacent nodes with connection coefficients less than a preset threshold are deleted, and nodes without connections are removed to obtain an updated causal network. Specifically, the preset threshold is set to 0.1, but the implementer can adjust it as needed. Edges (connections) between adjacent nodes with small connection coefficients are deleted, and nodes without connections in the causal network are removed to obtain an updated causal network.

[0056] Step S4: Based on the current monitoring information of each indicator of the current patient during postoperative care, and the connection coefficients between adjacent nodes in the updated causal network, obtain the risk prediction probability of each postoperative complication of the current patient.

[0057] After updating the causal network of complications, we can further combine the postoperative nursing monitoring information of the current patient for each indicator to conduct a preliminary analysis of the causal chains that may cause complications in the current patient. For example, if the current patient has a problem with untimely changes of incision dressings, it will be easier to screen out the causal chains that the current patient may be associated with in the causal network. The connection coefficient between adjacent nodes in each causal chain initially reflects the risk association or influence between the node and the complication, which can help prepare for the subsequent prediction of the risk probability of the current patient's corresponding complications based on the causal chains.

[0058] It should be noted that the prediction of postoperative complications for the current patient is carried out in real time during the postoperative care process. In one embodiment of the present invention, the analysis and description are based on any time point. The method for obtaining the current monitoring information of each indicator is the same as the method for obtaining historical patient information in step S1, and will not be repeated here.

[0059] Preferably, in one embodiment of the present invention, the method for obtaining the risk prediction probability includes: Based on the current monitoring information of each indicator during the postoperative care of the current patient, all reference causal chains of the current patient are determined in the updated causal network; the risk prediction probability of the corresponding complication is obtained according to the connection coefficient between adjacent nodes in each reference causal chain.

[0060] Specifically, the current monitoring information of each indicator of the current patient during postoperative care is analyzed. For example, if the patient's body temperature shows a downward trend during postoperative care, and the change of incision dressing is delayed or missed, it may mean that the patient's vital signs have changed, immunity may have decreased, and incision exudation may occur, which may lead to infection. Then, the corresponding nodes such as body temperature, incision dressing change and incision exudation can be screened in the causal network, and all causal chains to which these nodes belong can be determined. These causal chains are used as the reference causal chains for the current patient, that is, the risk paths that may lead to the corresponding complications. In a preferred embodiment of the present invention, obtaining the predicted risk probability of the corresponding complication based on the connection coefficient between adjacent nodes in each reference causal chain includes: The cumulative product of the connection coefficients between all adjacent nodes in each reference causal chain is normalized to obtain the risk prediction probability of the corresponding complication.

[0061] It should be noted that the cumulative result is mapped to the sigmoid function for normalization. When the complications corresponding to different reference causal chains are the same, the risk prediction probabilities of the complications corresponding to the reference causal chains of the same complication are averaged to obtain the final risk prediction probability.

[0062] In another embodiment of the present invention, after obtaining the reference causal chain, the implementer can also obtain the risk probability of each node based on forward reasoning of the Bayesian network, which is a well-known technology and will not be described in detail here; then in each reference causal chain, the ratio of the risk probability of the subsequent node to the risk probability of the preceding node between adjacent nodes is used as the risk amplification parameter between adjacent nodes; then the cumulative product of the risk amplification parameters between all adjacent nodes in each reference causal chain is normalized to obtain the risk amplification weight; the product of the risk amplification weight and the risk probability of the first node in the reference causal chain is used as the risk prediction probability of the corresponding complication. The implementer can then issue risk warnings and further simulate the intervention effects of different intervention measures to provide nursing staff with nursing references to reduce the risk of complications; for example, for the causal chain of "insufficient activity → slow blood flow → thrombosis", two interventions can be simulated: ① Increase the duration of time spent out of bed (intervention feature A - insufficient activity); ② Use of anticoagulants (intervention feature B - slow blood flow); Calculate the risk change for each intervention: For example, increasing activity reduces the probability of insufficient activity from 20% to 5%, with a risk amplification weight of 4. Then, the risk of complications before the intervention = 20% × 4 = 80%; the risk of complications after the intervention = 5% × 4 = 20%, a risk reduction of 60%. Similarly, the risk change before and after intervention with anticoagulants can be calculated. By comparing the risk reduction of different interventions, the optimal solution is automatically ranked. If ① is better than ②, then the risk of thrombosis is reduced by increasing activity. At the same time, individual patient differences are taken into account. For example, elderly patients may have poor tolerance to increased activity, and the intensity of intervention needs to be adjusted according to their physical characteristics (such as preoperative activity ability).

[0063] Thus, the risk of complications in the current postoperative care process for patients has been predicted.

[0064] In summary, this invention first analyzes the correlation coefficients between each indicator and each complication based on a large amount of postoperative care information from historical patients, and constructs a causal network of complications. Then, it analyzes the changing correlations between the monitoring information of adjacent nodes' corresponding indicators in each causal chain, as well as the correlation coefficients between the indicator corresponding to each node and the complication corresponding to the last node, obtaining the connection coefficients between adjacent nodes and updating the causal network. Finally, based on the current monitoring information of each indicator during the current patient's postoperative care, it predicts the risk of each postoperative complication for the current patient. This invention, based on postoperative care references provided by a large number of historical patients, constructs a risk-driven causal network that accurately reflects different complications. Therefore, it can accurately predict the risk of postoperative complications by leveraging the risk transmission causal chains of complications within the causal network, thus improving the effectiveness of risk prediction.

[0065] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0066] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for predicting the risk of postoperative complications in surgical care, characterized in that, The method includes: Acquire monitoring information for each indicator during postoperative care for each historical patient, as well as the timing of the occurrence of each complication; Based on the temporal distribution of the occurrence time of each complication in all historical patients, and the distribution of abnormal deviations in the temporal data of each indicator, the correlation coefficient between each indicator and each complication is obtained. A causal network of complications is constructed based on the monitoring information of all historical patients' corresponding indicators and complications. The last node of each causal chain in the causal network is a complication, and each non-last node corresponds to at least one indicator. The connection coefficient between adjacent nodes is obtained and the causal network is updated based on the change correlation between the monitoring information of the indicators corresponding to adjacent nodes in each causal chain and the correlation coefficient between the indicator corresponding to each node and the complication corresponding to the last node. Based on the current monitoring information of each indicator of the current patient during postoperative care, and the connection coefficients between adjacent nodes in the updated causal network, the risk prediction probability of each postoperative complication of the current patient is obtained.

2. The method for predicting the risk of postoperative complications in surgical care according to claim 1, characterized in that, The method for obtaining the correlation coefficient includes: The postoperative care process is divided into monitoring sub-periods of preset duration; For each complication, the frequency of occurrence of the complication within each monitoring sub-period is determined based on the time of occurrence of the complication in all historical patients, and a distribution sequence of the occurrence of the complication is constructed based on the frequency of occurrence within each monitoring sub-period. For each indicator, in each monitoring sub-period, based on the deviation of the concentrated characteristics of the monitoring information of each historical patient relative to the corresponding preset standard, the abnormal deviation parameter of the indicator is obtained, and the abnormal distribution sequence of the indicator is constructed based on the abnormal deviation parameter of the indicators of all historical patients in each monitoring sub-period. Based on the correlation between the abnormal distribution sequence of each indicator and the occurrence distribution sequence of each complication, the correlation coefficient between each indicator and each complication is obtained.

3. The method for predicting the risk of postoperative complications in surgical care according to claim 2, characterized in that, The method for obtaining the abnormal distribution sequence includes: For each indicator, in each monitoring sub-period, historical patients whose abnormal deviation parameter is greater than the corresponding preset threshold are regarded as abnormal patients, and the frequency of occurrence of abnormal patients in each monitoring sub-period is used as sequence elements to construct an abnormal distribution sequence.

4. The method for predicting the risk of postoperative complications in surgical care according to claim 2, characterized in that, The method for obtaining the connection coefficients includes: On each causal chain, the connection rigidity coefficient between adjacent nodes is obtained based on the change correlation between the monitoring information of the corresponding indicators of adjacent nodes and the difference in the number of corresponding indicators of adjacent nodes. In each causal chain, the average of the correlation coefficients between each indicator corresponding to each node and the complication corresponding to the last node is used as the node weight of each node; the connection weight between adjacent nodes is obtained based on the difference in the node weights of adjacent nodes. The connection rigidity coefficient is weighted using the connection weights, and the normalized value of the weighted result is used as the connection coefficient between the corresponding adjacent nodes.

5. The method for predicting the risk of postoperative complications in surgical care according to claim 4, characterized in that, The method for obtaining the connection stiffness coefficient includes: For each adjacent node, the indicator corresponding to the preceding node is taken as the target indicator, and the indicator corresponding to the following node is taken as the reference indicator. Based on the set characteristics of the Pearson correlation coefficient between the abnormal distribution sequences corresponding to each target indicator and each reference indicator, the rigid parameters between adjacent nodes are obtained. The ratio of the number of target indicators to the number of reference indicators is taken as the rigid weight between adjacent nodes, where the number of target indicators is the numerator. The rigid parameters are weighted using the rigid weight, and the weighted result is taken as the connection rigidity coefficient between the corresponding adjacent nodes.

6. The method for predicting the risk of postoperative complications in surgical care according to claim 4, characterized in that, The method for obtaining the connection weights includes: In adjacent nodes, the ratio of the node weight of the preceding node to the node weight of the following node is used as the connection weight between adjacent nodes, where the node weight of the preceding node is the denominator.

7. The method for predicting the risk of postoperative complications in surgical care according to claim 1, characterized in that, Methods for updating causal networks include: In a causal network, edges between adjacent nodes whose connection coefficients are less than a preset threshold are deleted, and nodes with no connection are removed to obtain an updated causal network.

8. The method for predicting the risk of postoperative complications in surgical care according to claim 1, characterized in that, The method for obtaining the risk prediction probability includes: Based on the current monitoring information of each indicator during the postoperative care of the current patient, all reference causal chains of the current patient are determined in the updated causal network; the risk prediction probability of the corresponding complication is obtained according to the connection coefficient between adjacent nodes in each reference causal chain.

9. A method for predicting the risk of postoperative complications in surgical care according to claim 8, characterized in that, The risk prediction probability of the corresponding complication is obtained based on the connection coefficient between adjacent nodes in each reference causal chain, including: The cumulative product of the connection coefficients between all adjacent nodes in each reference causal chain is normalized to obtain the risk prediction probability of the corresponding complication.

10. A method for predicting the risk of postoperative complications in surgical care according to claim 1, characterized in that, The indicators include at least each vital sign and each nursing procedure.

Images