Methods, devices, equipment, storage media and products for improving alarm accuracy

Abstract

This invention provides a method, apparatus, device, storage medium, and product for improving alarm accuracy, belonging to the field of healthcare information processing technology. The method includes: acquiring a historical sample set, a real-time vital sign dataset, and real-time prediction results for identifying an early warning model; calculating the reliability of the real-time prediction results; generating vital sign weights when the reliability of the real-time prediction results is less than or equal to a preset reliability threshold; correcting the vital sign data in the real-time vital sign dataset according to the vital sign weights, generating a corrected vital sign dataset and a corrected risk probability; outputting an alarm signal when the corrected risk probability is greater than a preset probability threshold, and outputting a silent signal when the corrected risk probability is less than or equal to the preset probability threshold. The alarm accuracy improvement method, apparatus, device, storage medium, and product described in this invention can improve the alarm accuracy of an EHR system and reduce the probability of false alarms and missed alarms.

Description

Technical Field

[0001] This invention belongs to the field of medical and health care information processing technology, and in particular relates to a method, device, equipment, storage medium and product for improving alarm accuracy. Background Technology

[0002] An EHR (Electronic Health Records) system is a system specifically designed to record, store, and manage patient health data. Compared to traditional paper-based medical records, EHR systems can automatically collect patients' vital signs data through various devices and analyze this data to provide medical decision support for healthcare professionals, enabling them to develop more accurate and timely treatment plans.

[0003] In existing technologies, EHR systems typically incorporate an identification and early warning model. After collecting a patient's vital signs data, the model predicts the patient's health risk probability based on this data and issues an alarm when the probability exceeds a preset threshold, thus alerting healthcare professionals to pay attention and make timely clinical decisions.

[0004] However, due to limitations in the predictive accuracy of the identification and early warning model, EHR systems inevitably experience false alarms and missed alarms. In practical applications, false alarms occur when the EHR system issues an alarm even though the patient is not in a critical condition. Frequent occurrences of this increase the workload of medical staff and reduce their trust in the EHR system. Missed alarms occur when a patient is in a critical condition but the EHR system fails to issue an alarm, resulting in the patient not receiving timely treatment and causing serious adverse effects. To address this issue, those skilled in the art typically choose to build more complex deep learning models or increase the number of parameters to improve the predictive accuracy of the identification and early warning model. However, such solutions significantly increase computational and data acquisition costs, thus failing to meet practical application needs. Summary of the Invention

[0005] In view of this, the present invention aims to provide a method, apparatus, device, storage medium and product for improving alarm accuracy, so as to solve the above-mentioned technical problems.

[0006] To achieve the above objectives, the technical solution created by this invention is implemented as follows:

[0007] In a first aspect, embodiments of the present invention provide a method for improving alarm accuracy, including:

[0008] Acquire the historical sample set, real-time vital sign dataset, and real-time prediction results of the identification and early warning model. The historical sample set includes multiple historical prediction results and the historical vital sign dataset corresponding to each historical prediction result.

[0009] The reliability of real-time prediction results is calculated based on real-time vital signs datasets and historical sample sets.

[0010] When the confidence level of the real-time prediction result is less than or equal to the preset confidence level threshold, the importance of each vital sign data in the historical vital sign dataset is determined based on the historical sample set, and the vital sign weight of the vital sign data is generated based on the importance level.

[0011] The vital signs data in the real-time vital signs dataset are corrected according to the vital signs weights to generate a corrected vital signs dataset. The corrected vital signs dataset is then input into the identification and early warning model to generate a corrected risk probability.

[0012] When the corrected risk probability is greater than the preset probability threshold, a high-risk corrected prediction result is generated and an alarm signal is output. When the corrected risk probability is less than or equal to the preset probability threshold, a low-risk corrected prediction result is generated and a silent signal is output.

[0013] Furthermore, the real-time prediction results include high-risk real-time prediction results and low-risk real-time prediction results;

[0014] The historical sample set includes a high-risk historical sample subset and a low-risk historical sample subset;

[0015] The high-risk historical sample subset includes high-risk historical prediction results and the corresponding historical vital sign datasets;

[0016] The low-risk historical sample subset includes low-risk historical prediction results and the corresponding historical vital sign datasets;

[0017] The calculation of the reliability of real-time prediction results based on real-time vital sign datasets and historical sample sets includes:

[0018] Calculate the consistency metric for each historical vital sign dataset in the high-risk historical sample subset, and sort the consistency metrics of each historical vital sign dataset in ascending order to generate a high-risk historical consistency list.

[0019] Calculate the consistency metric for each historical vital sign dataset in the low-risk historical sample subset, and sort the consistency metrics of each historical vital sign dataset in ascending order to generate a low-risk historical consistency list.

[0020] When the real-time prediction result is a high-risk real-time prediction result, the real-time vital signs dataset is used as a new vital signs dataset in the high-risk historical sample subset. The consistency metric of the new vital signs dataset is calculated, and the credibility of the real-time prediction result is determined according to the sorting position of the consistency metric of the new vital signs dataset in the high-risk historical consistency list.

[0021] When the real-time prediction result is a low-risk real-time prediction result, the real-time vital signs dataset is used as a new vital signs dataset in the low-risk historical sample subset. The consistency metric of the new vital signs dataset is calculated, and the credibility of the real-time prediction result is determined according to the sorting position of the consistency metric of the new vital signs dataset in the low-risk historical consistency list.

[0022] Furthermore, before calculating the reliability of the real-time prediction result based on the real-time vital sign dataset and the historical sample set, the alarm accuracy improvement method further includes:

[0023] Determine the normal range of historical vital signs dataset based on historical sample sets;

[0024] The direction and rate of change of the real-time vital signs dataset within a preset time period are obtained. If the real-time vital signs dataset changes towards the normal range, the preset confidence threshold is lowered and adjusted, and the step size of the lowering of the preset confidence threshold is inversely proportional to the rate of change. If the real-time vital signs dataset changes away from the normal range, the preset confidence threshold is raised and adjusted, and the step size of the raising of the preset confidence threshold is directly proportional to the rate of change.

[0025] Furthermore, before calculating the reliability of the real-time prediction result based on the real-time vital sign dataset and the historical sample set, the alarm accuracy improvement method further includes:

[0026] The false alarm rate and missed alarm rate of the EHR system are obtained. When the false alarm rate increases, the preset confidence threshold is increased. When the missed alarm rate increases, the preset confidence threshold is decreased.

[0027] Furthermore, after calculating the reliability of the real-time prediction results based on the real-time vital signs dataset and the historical sample set, the alarm accuracy improvement method further includes:

[0028] When the credibility of the real-time prediction result is greater than the preset credibility threshold, the type of the real-time prediction result is determined. If the real-time prediction result is a high-risk real-time prediction result, an alarm signal is output. If the real-time prediction result is a low-risk real-time prediction result, a silence signal is output.

[0029] Furthermore, before correcting the vital sign data in the real-time vital sign dataset according to the vital sign weights, the alarm accuracy improvement method further includes:

[0030] Determine the normal range of historical vital signs dataset based on historical sample sets;

[0031] The system obtains the object attribution of the real-time vital signs dataset and the direction of change of the real-time vital signs dataset within a preset time period. If the real-time vital signs dataset changes towards the normal range, the weight of the vital signs is reduced. The reduction adjustment step size when the real-time vital signs dataset belongs to a high-risk object is smaller than the reduction adjustment step size when the real-time vital signs dataset belongs to a normal object. If the real-time vital signs dataset changes away from the normal range, the weight of the vital signs is increased. The increase adjustment step size when the real-time vital signs dataset belongs to a high-risk object is larger than the increase adjustment step size when the real-time vital signs dataset belongs to a normal object.

[0032] Secondly, embodiments of the present invention also provide an alarm accuracy improvement device, comprising:

[0033] The acquisition module is used to acquire the historical sample set, real-time vital sign dataset and real-time prediction results of the identification and early warning model. The historical sample set includes multiple historical prediction results and the historical vital sign dataset corresponding to each historical prediction result.

[0034] The calculation module is used to calculate the reliability of real-time prediction results based on real-time vital signs datasets and historical sample sets;

[0035] The generation module is used to determine the importance of each vital sign data in the historical vital sign dataset based on the historical sample set when the confidence of the real-time prediction result is less than or equal to a preset confidence threshold, and to generate the vital sign weight of the vital sign data based on the importance.

[0036] The correction module is used to correct the vital sign data in the real-time vital sign dataset according to the vital sign weight, generate a corrected vital sign dataset, and input the corrected vital sign dataset into the identification and early warning model to generate the corrected risk probability.

[0037] The output module is used to generate a high-risk correction prediction result and output an alarm signal when the corrected risk probability is greater than a preset probability threshold, and to generate a low-risk correction prediction result and output a silent signal when the corrected risk probability is less than or equal to the preset probability threshold.

[0038] Thirdly, embodiments of the present invention also provide an apparatus, comprising:

[0039] One or more processors;

[0040] Storage device for storing one or more programs;

[0041] When the one or more programs are executed by the one or more processors, the one or more processors implement the alarm accuracy improvement method provided in any of the above embodiments.

[0042] Fourthly, embodiments of the present invention also provide a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform the alarm accuracy improvement method provided in any of the above embodiments.

[0043] Fifthly, embodiments of the present invention also provide a computer program product, including a computer program that, when executed by a processor, implements the alarm accuracy improvement method provided in any of the above embodiments.

[0044] Compared with existing technologies, the alarm accuracy improvement method, apparatus, device, storage medium, and product described in this invention have the following advantages:

[0045] This invention provides a method, apparatus, device, storage medium, and product for improving alarm accuracy. It calculates the reliability of real-time prediction results based on real-time vital sign datasets and historical sample sets. When the reliability of the real-time prediction result is less than or equal to a preset reliability threshold, it generates vital sign weights based on the importance of each vital sign data point and corrects the real-time vital sign dataset according to these weights, thereby generating a corrected vital sign dataset. Subsequently, the corrected vital sign dataset is input into an identification and early warning model to generate a corrected risk probability. The corrected risk probability is compared with a preset probability threshold to generate a corrected prediction result, and an alarm signal or a silence signal is output accordingly. Compared with existing technologies, this invention avoids significantly increasing computational and data acquisition costs and, through vital sign weights, makes the corrected risk probability more sensitive, thus enabling the prediction results to more accurately reflect the patient's actual condition, thereby improving the alarm accuracy of the EHR system and reducing the probability of false alarms and missed alarms. Attached Figure Description

[0046] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0047] Figure 1 A flowchart of the alarm accuracy improvement method described in Embodiment 1 of the present invention is provided;

[0048] Figure 2 A flowchart of the alarm accuracy improvement method described in Embodiment 2 of this invention is provided;

[0049] Figure 3 A flowchart of the alarm accuracy improvement method described in Embodiment 3 of the present invention is provided;

[0050] Figure 4 This invention provides a schematic diagram of the alarm accuracy improvement device described in Embodiment 4 of the present invention.

[0051] Figure 5 The structural diagram of the device described in Embodiment 5 of the present invention is shown. Detailed Implementation

[0052] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0053] Example 1

[0054] Figure 1 The flowchart of the alarm accuracy improvement method provided in Embodiment 1 of the present invention specifically includes the following steps:

[0055] Step 110: Obtain the historical sample set, real-time vital sign dataset, and real-time prediction results of the identification and early warning model. The historical sample set includes multiple historical prediction results and the historical vital sign dataset corresponding to each historical prediction result.

[0056] When the EHR system is in operation, the data acquisition device continuously collects the patient's vital signs data (such as heart rate, blood pressure, blood oxygen saturation, and blood glucose), and inputs this data into the identification and early warning model. The model then predicts the patient's health risks based on the vital signs data, enabling timely alerts when health risks are identified, prompting medical staff to make clinical decisions. Correspondingly, the EHR system also stores the historical input data and output results of the identification and early warning model for record-keeping and subsequent management.

[0057] To improve the alarm accuracy of the EHR system, this embodiment will acquire the historical sample set, real-time vital sign dataset, and real-time prediction results of the identification and early warning model. The historical sample set should include multiple historical prediction results (i.e., historical output results) and the corresponding historical vital sign dataset (i.e., historical input data) for each historical prediction result. The real-time vital sign dataset is the real-time input data of the identification and early warning model, and the real-time prediction results are the real-time output results generated by the identification and early warning model based on the real-time input data.

[0058] In the subsequent work of the EHR system, the historical sample set will be used to calculate the confidence level of subsequent prediction results, thereby avoiding false alarms and missed alarms caused by prediction results with low confidence.

[0059] It should be noted that, since the identification and early warning model typically requires input of multiple types of vital sign data when predicting a patient's health risk, and the sampling frequencies of the data acquisition devices used to collect different types of vital sign data differ, those skilled in the art usually set the sampling period according to the actual situation. This allows multiple vital sign data to form a vital sign dataset within the same sampling period and be input into the identification and early warning model, facilitating the model's output of prediction results. Accordingly, the historical vital sign dataset in this embodiment should include multiple vital sign data within the same historical sampling period, while the real-time vital sign dataset should include multiple vital sign data within the same real-time sampling period.

[0060] Step 120: Calculate the reliability of the real-time prediction results based on the real-time vital signs dataset and the historical sample set.

[0061] After acquiring historical sample sets, real-time vital signs datasets, and real-time prediction results, this embodiment calculates the reliability of the real-time prediction results and compares the calculation result with a preset reliability threshold to determine the subsequent processing method. Specifically, when the reliability of the real-time prediction result is greater than the preset reliability threshold, it proves that the real-time prediction result can accurately reflect the patient's actual situation. In this case, it can be determined whether to trigger an alarm based on the real-time prediction result. When the reliability of the real-time prediction result is less than or equal to the preset reliability threshold, it proves that the real-time prediction result may have errors. In this case, subsequent processing should be performed on the real-time prediction result to improve the alarm accuracy of the EHR system.

[0062] Optionally, the real-time prediction results in this embodiment may include two types: high-risk real-time prediction results and low-risk real-time prediction results. High-risk real-time prediction results indicate that the patient currently has a significant health risk and requires triggering an alarm to remind medical staff to take timely clinical measures. Low-risk real-time prediction results indicate that the patient currently has a low health risk and do not require triggering an alarm.

[0063] The historical sample set can also be divided into a high-risk historical sample subset and a low-risk historical sample subset based on the type of historical prediction results. The high-risk historical sample subset includes high-risk historical prediction results (i.e., historical prediction results that have triggered alarms in the historical record) and the corresponding historical vital sign datasets. The low-risk historical sample subset includes low-risk historical prediction results (i.e., historical prediction results that have not triggered alarms in the historical record) and the corresponding historical vital sign datasets.

[0064] Accordingly, the reliability of real-time prediction results calculated based on real-time vital sign datasets and historical sample sets can be specifically optimized as follows:

[0065] Calculate the consistency metric for each historical vital sign dataset in the high-risk historical sample subset, and sort the consistency metrics of each historical vital sign dataset in ascending order to generate a high-risk historical consistency list.

[0066] Calculate the consistency metric for each historical vital sign dataset in the low-risk historical sample subset, and sort the consistency metrics of each historical vital sign dataset in ascending order to generate a low-risk historical consistency list.

[0067] When the real-time prediction result is a high-risk real-time prediction result, the real-time vital signs dataset is used as a new vital signs dataset in the high-risk historical sample subset. The consistency metric of the new vital signs dataset is calculated, and the credibility of the real-time prediction result is determined according to the sorting position of the consistency metric of the new vital signs dataset in the high-risk historical consistency list.

[0068] When the real-time prediction result is a low-risk real-time prediction result, the real-time vital signs dataset is used as a new vital signs dataset in the low-risk historical sample subset. The consistency metric of the new vital signs dataset is calculated, and the credibility of the real-time prediction result is determined according to the sorting position of the consistency metric of the new vital signs dataset in the low-risk historical consistency list.

[0069] In existing technologies, clustering algorithms are typically used to determine the reliability of prediction results. This involves assessing the reliability of real-time data based on the similarity between real-time and historical data; the higher the similarity, the higher the reliability. However, due to limitations in the prediction accuracy of early warning models, historical predictions in the historical sample set may contain errors. If traditional clustering algorithms are used to calculate the reliability of real-time predictions, the reliability of the real-time predictions will also be subject to errors, thus compromising the accuracy of subsequent processing.

[0070] Therefore, this embodiment will first calculate the consistency metric value for each historical vital sign dataset in the high-risk sample subset and the low-risk sample subset, thereby using the consistency metric value to judge the credibility of the historical sample set. It should be noted that the consistency metric value is used to characterize the degree of similarity between different input vital sign datasets when the identification and early warning model outputs the same type of prediction results, and the higher the similarity, the more credible the historical prediction results generated based on the historical vital sign dataset.

[0071] The following example illustrates the specific calculation method for the consistency metric in this embodiment, using the calculation of a consistency metric value for a certain historical vital sign dataset within a high-risk historical sample subset as an example:

[0072] The first step is to calculate the minimum Euclidean distance A from the historical vital signs dataset to all historical vital signs datasets in the low-risk historical sample subset;

[0073] The second step is to calculate the minimum Euclidean distance B between this historical vital sign dataset and other historical vital sign datasets in the high-risk historical sample subset;

[0074] The third step is to calculate the ratio C of A to B, and use C as a consistency measure of the historical vital signs dataset.

[0075] Since Euclidean distance can represent the true distance between two points in space, a larger ratio C indicates that the historical vital sign dataset is closer to the high-risk historical sample subset, meaning that the historical vital sign dataset has a higher similarity to other historical vital sign datasets in the high-risk historical sample subset. Conversely, a smaller ratio C indicates that the historical vital sign dataset is closer to the low-risk historical sample subset, meaning that the historical vital sign dataset has a higher similarity to historical vital sign datasets in the low-risk historical sample subset.

[0076] After calculating the consistency metric values ​​for the historical vital sign dataset, this embodiment sorts the consistency metric values ​​in ascending order to generate a high-risk historical consistency list and a low-risk historical consistency list. Since the consistency metric values ​​are higher when ranked later in the consistency list, a later ranking of a certain consistency metric value indicates higher reliability of the prediction results generated from the corresponding vital sign dataset. Similarly, after treating the real-time vital sign dataset as a new dataset based on the type of real-time prediction result and calculating the consistency metric values ​​for the new dataset, the reliability of the real-time prediction results can be determined by the ranking of the consistency metric values ​​of the new dataset in the corresponding consistency list.

[0077] To facilitate the conversion of the ranking position of the consistency metric values ​​of newly added vital sign datasets in the corresponding consistency list into specific confidence scores, this embodiment will use a high-risk real-time prediction result as an example to illustrate the conversion method for confidence scores:

[0078] Step 1: After calculating the consistency metric of the newly added vital signs dataset, substitute the consistency metric of the newly added vital signs dataset into the high-risk historical consistency list, and obtain the position number X1 of the consistency metric of the newly added vital signs dataset.

[0079] Step 2: Obtain the position number X2 of the consistency metric value of the last historical vital sign dataset in the high-risk historical consistency list;

[0080] Step 3: First, calculate the ratio Y of X1 to X2, and use the ratio Y as the confidence level of the prediction result (i.e., the real-time prediction result) corresponding to the newly added vital signs dataset (i.e., the real-time vital signs dataset).

[0081] For example, suppose the high-risk historical consistency list contains consistency metrics for 100 historical vital sign datasets. In this case, the position index X2 of the consistency metric for the last historical vital sign dataset is 100. If the consistency metric for a newly added vital sign dataset is ranked 70th in the high-risk historical consistency list, then its position index X1 is 70. The ratio Y is then 70 / 100, and therefore the confidence level of its prediction result is 0.7.

[0082] As an optional implementation of this embodiment, when the credibility of the real-time prediction result is greater than a preset credibility threshold, after calculating the credibility of the real-time prediction result based on the real-time vital signs dataset and the historical sample set, this embodiment may add the following steps:

[0083] When the credibility of the real-time prediction result is greater than the preset credibility threshold, the type of the real-time prediction result is determined. If the real-time prediction result is a high-risk real-time prediction result, an alarm signal is output. If the real-time prediction result is a low-risk real-time prediction result, a silence signal is output.

[0084] When the reliability of the real-time prediction result exceeds a preset reliability threshold, it proves that the real-time prediction result accurately reflects the patient's actual situation. If the real-time prediction result is high-risk, it proves that the patient needs timely treatment, and an alarm signal should be output to prompt the EHR system to alert promptly. If the real-time prediction result is low-risk, it proves that the patient's current health risk is low, and a silent signal can be output to reduce the alarm frequency of the EHR system.

[0085] Step 130: When the confidence level of the real-time prediction result is less than or equal to the preset confidence level threshold, determine the importance of each vital sign data in the historical vital sign dataset based on the historical sample set, and generate the vital sign weight of the vital sign data based on the importance level.

[0086] When the reliability of a real-time prediction result is less than or equal to a preset reliability threshold, it indicates that the real-time prediction result may contain errors. If the type of the real-time prediction result is directly used to determine whether to trigger an alarm, it will lead to frequent false alarms or missed alarms in the EHR system. To solve this problem, this embodiment will determine the importance of each vital sign data in the historical vital sign dataset based on the historical sample set, and generate vital sign weights based on the importance of the vital sign data. This vital sign weighting will improve the accuracy and sensitivity of the prediction results of the identification and early warning model, so as to more accurately reflect the actual situation of the patient and improve the alarm accuracy of the EHR system.

[0087] It's important to note that the importance of vital sign data refers to the significance of the corresponding vital sign indicator in predicting a patient's health risk of a certain type of disease. For example, when predicting a patient's health risk of cardiovascular disease, blood pressure and heart rate are two commonly used vital sign indicators in this field, with blood pressure being more important than heart rate. Therefore, when generating the weights of vital sign data based on importance, the weight of the vital sign data reflecting blood pressure will be greater than the weight of the vital sign data reflecting heart rate, thus making the subsequent prediction process more sensitive and accurate in predicting the health risk of cardiovascular disease.

[0088] Step 140: Correct the vital sign data in the real-time vital sign dataset according to the vital sign weights to generate a corrected vital sign dataset, and input the corrected vital sign dataset into the identification and early warning model to generate the corrected risk probability.

[0089] After generating the vital sign weights, this embodiment corrects the vital sign data in the real-time vital sign dataset according to the weights, thereby generating a corrected vital sign dataset. This corrected vital sign dataset is then input into the identification and early warning model to generate a corrected risk probability. Since the vital sign weights reflect the importance of vital sign data in predicting disease and health risks, the corrected risk probability generated by inputting the corrected vital sign dataset into the identification and early warning model has better accuracy and sensitivity for disease-related health risks, more accurately reflecting the patient's actual condition and thus improving the alarm accuracy of the EHR system.

[0090] Step 150: When the corrected risk probability is greater than the preset probability threshold, generate a high-risk corrected prediction result and output an alarm signal; when the corrected risk probability is less than or equal to the preset probability threshold, generate a low-risk corrected prediction result and output a silent signal.

[0091] After obtaining the corrected risk probability, this embodiment compares the corrected risk probability with a preset probability threshold to facilitate the determination of the prediction result type corresponding to the corrected risk probability. When the corrected risk probability is greater than the preset probability threshold, it indicates that the patient has a high disease-related health risk and requires timely treatment. In this case, a high-risk corrected prediction result should be generated, and an alarm signal should be output to trigger the EHR system. When the corrected risk probability is less than or equal to the preset probability threshold, it indicates that the patient has a low disease-related health risk. In this case, a low-risk corrected prediction result should be generated, and a silent signal should be output to avoid triggering the EHR system.

[0092] This embodiment can calculate the reliability of real-time prediction results based on real-time vital sign datasets and historical sample sets. When the reliability of the real-time prediction results is less than or equal to a preset reliability threshold, it can generate vital sign weights based on the importance of each vital sign data point, and then correct the real-time vital sign dataset according to the vital sign weights, thereby generating a corrected vital sign dataset. Subsequently, the corrected vital sign dataset is input into the identification and early warning model to generate a corrected risk probability, and a corrected prediction result is generated by comparing the corrected risk probability with a preset probability threshold. An alarm signal or a silence signal is then output according to the corrected prediction result. Therefore, it can avoid significantly increasing computational and data acquisition costs, and can make the corrected risk probability more sensitive through vital sign weights, thereby making the prediction results more accurately reflect the patient's actual situation, thus improving the alarm accuracy of the EHR system and reducing the probability of false alarms and missed alarms.

[0093] As another optional implementation of this embodiment, before calculating the reliability of the real-time prediction result based on the real-time vital signs dataset and the historical sample set, the alarm accuracy improvement method may further include the following steps:

[0094] The false alarm rate and missed alarm rate of the EHR system are obtained. When the false alarm rate increases, the preset confidence threshold is increased. When the missed alarm rate increases, the preset confidence threshold is decreased.

[0095] During normal operations, healthcare staff respond to alarms from the Emergency Health Response (EHR) system by proceeding to the patient's area to provide clinical care. After the care is completed, they record the patient's condition in the EHR system. Similarly, when the EHR system is not alarming, healthcare staff regularly conduct ward rounds and rounds to understand the patient's condition and record the results in the EHR system. Therefore, the false alarm rate and missed alarm rate of the EHR system can be determined based on the records recorded by healthcare staff. When the false alarm rate increases, raising the preset confidence threshold allows for more accurate prediction of whether an alarm has been triggered. Conversely, when the missed alarm rate increases, lowering the preset confidence threshold allows for more accurate prediction of whether an alarm has been triggered, thus improving the practical application effectiveness of this method.

[0096] Example 2

[0097] Figure 2 This is a flowchart of the alarm accuracy improvement method provided in Embodiment 2 of the present invention. This embodiment is an optimization based on the above embodiment. In this embodiment, before calculating the reliability of the real-time prediction result based on the real-time vital sign dataset and historical sample set, the following steps can be added:

[0098] The normal range of the historical vital signs dataset is determined based on the historical sample set; the direction and rate of change of the real-time vital signs dataset within a preset time period are obtained; if the real-time vital signs dataset changes towards the normal range, the preset confidence threshold is lowered, and the step size of the lowering of the preset confidence threshold is inversely proportional to the rate of change; if the real-time vital signs dataset changes away from the normal range, the preset confidence threshold is raised, and the step size of the raising of the preset confidence threshold is directly proportional to the rate of change.

[0099] Accordingly, the alarm accuracy improvement method provided in this embodiment specifically includes:

[0100] Step 210: Obtain the historical sample set, real-time vital sign dataset, and real-time prediction results of the identification and early warning model. The historical sample set includes multiple historical prediction results and the historical vital sign dataset corresponding to each historical prediction result.

[0101] Step 220: Determine the normal range of the historical vital signs dataset based on the historical sample set; obtain the direction and rate of change of the real-time vital signs dataset within a preset time period. If the real-time vital signs dataset changes towards the normal range, the preset confidence threshold is lowered, and the step size of the lowering of the preset confidence threshold is inversely proportional to the rate of change. If the real-time vital signs dataset changes away from the normal range, the preset confidence threshold is raised, and the step size of the raising of the preset confidence threshold is directly proportional to the rate of change.

[0102] Normally, a patient's vital signs will change with their physical condition. When a patient is healthy, their vital signs will fluctuate within a certain range, which is called the normal range for vital signs. When a patient is ill, their vital signs will deviate from the normal range, and the greater the deviation, the more severe the illness tends to be.

[0103] Therefore, this embodiment will determine the normal range of the historical vital signs dataset based on the historical sample set, in order to analyze the changes in the real-time vital signs dataset and thus determine the trend of changes in the patient's physical condition. It should be noted that since the patient's vital signs data will not trigger an alarm in the EHR system when it is within the normal range, the historical vital signs dataset should be selected from the low-risk historical sample subset when determining the normal range of the historical vital signs dataset.

[0104] Accordingly, to reflect the changes in the real-time vital signs dataset, this embodiment should also obtain the direction and rate of change of the real-time vital signs dataset within a preset time period. The direction of change in the real-time vital signs data can determine whether the patient's physical condition is trending towards deterioration or recovery, while the rate of change can determine the speed of deterioration and recovery, allowing for flexible adjustment of the preset confidence threshold.

[0105] When the real-time vital signs dataset shifts towards the normal range, it indicates that the patient's physical condition is trending towards recovery. At this point, the preset confidence threshold can be lowered. After lowering the preset confidence threshold, the EHR system will rely more on the type of real-time prediction results to determine whether to trigger an alarm, thus reducing the EHR system's computing power requirements.

[0106] When the real-time vital signs dataset deviates from the normal range, it indicates that the patient's condition is deteriorating. At this point, the preset confidence threshold can be increased. After increasing the preset confidence threshold, the EHR system will rely more on correcting the type of prediction results to determine whether to trigger an alarm, thus becoming more accurate and sensitive, allowing medical staff to promptly address patients whose conditions are worsening.

[0107] In addition, since patients' vital signs do not change rapidly in a short period of time under normal circumstances, a high rate of change in vital signs can prove that the patient's physical condition has changed abnormally.

[0108] When the real-time vital signs dataset changes towards the normal range, a high rate of change in the real-time vital signs dataset clearly does not conform to normal recovery patterns. In this case, the adjustment step size of the preset confidence threshold should be inversely proportional to the rate of change (i.e., the higher the rate of change, the smaller the adjustment step size of the confidence threshold), thereby slowly reducing the preset confidence threshold and preventing the EHR system from prematurely determining whether to trigger an alarm based on real-time prediction results.

[0109] When the real-time vital signs dataset changes away from the normal range, a high rate of change indicates a rapid deterioration in the patient's condition. In this case, the adjustment step size for increasing the preset confidence threshold should be proportional to the rate of change (i.e., the higher the rate of change, the larger the adjustment step size for decreasing the confidence threshold). This will quickly increase the preset confidence threshold, allowing the EHR system to determine whether to trigger an alarm more quickly and earlier by correcting the prediction results.

[0110] Step 230: Calculate the credibility of the real-time prediction results based on the real-time vital signs dataset and the historical sample set.

[0111] Step 240: When the confidence level of the real-time prediction result is less than or equal to the preset confidence level threshold, determine the importance of each type of vital sign data in the historical vital sign dataset based on the historical sample set, and generate the vital sign weight of the vital sign data based on the importance level.

[0112] Step 250: Correct the vital sign data in the real-time vital sign dataset according to the vital sign weights to generate a corrected vital sign dataset, and input the corrected vital sign dataset into the identification and early warning model to generate the corrected risk probability.

[0113] Step 260: When the corrected risk probability is greater than the preset probability threshold, generate a high-risk corrected prediction result and output an alarm signal; when the corrected risk probability is less than or equal to the preset probability threshold, generate a low-risk corrected prediction result and output a silent signal.

[0114] This embodiment adds the following steps before calculating the reliability of the real-time prediction result based on the real-time vital signs dataset and the historical sample set: determining the normal range of the historical vital signs dataset based on the historical sample set; obtaining the direction and rate of change of the real-time vital signs dataset within a preset time period; if the real-time vital signs dataset changes towards the normal range, the preset reliability threshold is lowered, and the step size of the lowering adjustment of the preset reliability threshold is proportional to the rate of change; if the real-time vital signs dataset changes away from the normal range, the preset reliability threshold is raised, and the step size of the raising adjustment of the preset reliability threshold is proportional to the rate of change. This allows the preset reliability threshold to be adjusted according to the direction and rate of change of the real-time vital signs dataset within a preset time period, thereby adapting to changes in the patient's physical condition.

[0115] Example 3

[0116] Figure 3 This is a flowchart of the alarm accuracy improvement method provided in Embodiment 3 of the present invention. This embodiment is an optimization based on the above embodiment. In this embodiment, before correcting the vital sign data in the real-time vital sign dataset according to the vital sign weights, the following steps can be added:

[0117] The normal range of the historical vital sign dataset is determined based on the historical sample set; the object affiliation of the real-time vital sign dataset and the direction of change of the real-time vital sign dataset within a preset time period are obtained. If the real-time vital sign dataset changes towards the normal range, the weight of the vital sign is reduced, and the reduction adjustment step size when the real-time vital sign dataset belongs to a high-risk object is smaller than the reduction adjustment step size when the real-time vital sign dataset belongs to a normal object. If the real-time vital sign dataset changes away from the normal range, the weight of the vital sign is increased, and the increase adjustment step size when the real-time vital sign dataset belongs to a high-risk object is larger than the increase adjustment step size when the real-time vital sign dataset belongs to a normal object.

[0118] Specifically, the alarm accuracy improvement method provided in this embodiment includes:

[0119] Step 310: Obtain the historical sample set, real-time vital sign dataset, and real-time prediction results of the identification and early warning model. The historical sample set includes multiple historical prediction results and the historical vital sign dataset corresponding to each historical prediction result.

[0120] Step 320: Calculate the credibility of the real-time prediction results based on the real-time vital signs dataset and the historical sample set.

[0121] Step 330: When the confidence level of the real-time prediction result is less than or equal to the preset confidence level threshold, determine the importance of each type of vital sign data in the historical vital sign dataset based on the historical sample set, and generate the vital sign weight of the vital sign data based on the importance level.

[0122] Step 340: Determine the normal range of the historical vital sign dataset based on the historical sample set; obtain the object affiliation of the real-time vital sign dataset and the direction of change of the real-time vital sign dataset within a preset time period. If the real-time vital sign dataset changes towards the normal range, the vital sign weight is reduced, and the reduction adjustment step size when the real-time vital sign dataset belongs to a high-risk object is smaller than the reduction adjustment step size when the real-time vital sign dataset belongs to a normal object. If the real-time vital sign dataset changes away from the normal range, the vital sign weight is increased, and the increase adjustment step size when the real-time vital sign dataset belongs to a high-risk object is larger than the increase adjustment step size when the real-time vital sign dataset belongs to a normal object.

[0123] To ensure that the correction of vital signs data adapts to changes in the patient's physical condition, this embodiment adjusts the weights of vital signs according to changes in the patient's physical condition. Similar to the above embodiment, this embodiment also determines the normal range of historical vital signs dataset based on historical sample sets to facilitate analysis of changes in real-time vital signs datasets, thereby determining the trend reflecting changes in the patient's physical condition.

[0124] When the real-time vital signs dataset changes towards the normal range, it indicates that the patient's physical condition is trending towards recovery. At this point, the weights of the vital signs can be reduced to decrease the disease sensitivity of the identification and early warning model when generating corrected prediction results, thereby reducing the probability of triggering alarms.

[0125] When the real-time vital signs dataset changes away from the normal range, it indicates that the patient's physical condition is deteriorating. At this time, the weight of the vital signs can be increased to improve the disease sensitivity of the identification and early warning model when generating corrected prediction results, so as to more accurately capture the deterioration of the patient's condition and issue an alarm in a timely manner.

[0126] In practical applications, due to differences in patients' specific physical conditions (such as age differences and differences in chronic disease history), patients can be roughly divided into two types: high-risk patients and ordinary patients. The adjustment step size of the vital signs weights will differ according to the patient type. High-risk patients refer to those with poor physical condition (e.g., those over 60 years old or under 12 years old, or those with long-term chronic diseases), while ordinary patients refer to those with good physical condition (e.g., those between 12 and 60 years old, with no history of chronic diseases).

[0127] Specifically, when adjusting the weights of vital signs, if the real-time vital signs dataset is classified as a high-risk subject, the adjustment step size for reducing the weights of vital signs should be smaller than the adjustment step size when the real-time vital signs dataset is classified as a normal subject. This is to ensure that the weights of vital signs for high-risk subjects are reduced more smoothly, thereby enabling high-risk subjects to maintain a high level of disease prediction sensitivity during the recovery process.

[0128] When adjusting the weight of vital signs, if the real-time vital signs dataset belongs to a high-risk group, the adjustment step size of the weight of vital signs should be larger than the adjustment step size when the real-time vital signs dataset belongs to a normal group. This is so that the weight of vital signs of high-risk groups can be increased quickly when their condition worsens, thereby issuing an alarm in a timely manner.

[0129] Step 350: Correct the vital sign data in the real-time vital sign dataset according to the vital sign weights to generate a corrected vital sign dataset, and input the corrected vital sign dataset into the identification and early warning model to generate the corrected risk probability.

[0130] Step 360: When the corrected risk probability is greater than the preset probability threshold, generate a high-risk corrected prediction result and output an alarm signal; when the corrected risk probability is less than or equal to the preset probability threshold, generate a low-risk corrected prediction result and output a silent signal.

[0131] This embodiment adds the following steps before correcting the vital sign data in the real-time vital sign dataset according to the vital sign weights: determining the normal range of the historical vital sign dataset based on the historical sample set; obtaining the object classification of the real-time vital sign dataset and the direction of change of the real-time vital sign dataset within a preset time period; if the real-time vital sign dataset changes towards the normal range, the vital sign weights are adjusted by decreasing the weights, with the decrease adjustment step size when the real-time vital sign dataset belongs to a high-risk object being smaller than the decrease adjustment step size when the real-time vital sign dataset belongs to a normal object being larger; if the real-time vital sign dataset changes away from the normal range, the vital sign weights are adjusted by increasing the weights, with the increase adjustment step size when the real-time vital sign dataset belongs to a high-risk object being larger than the increase adjustment step size when the real-time vital sign dataset belongs to a normal object being larger. This allows the vital sign weights to be adjusted according to the direction of change of the real-time vital sign data within a preset time period and the object classification, thereby adapting to changes in the patient's physical condition and differences in physical fitness.

[0132] Example 4

[0133] Figure 4 A schematic diagram of the alarm accuracy improvement device provided in Embodiment 4 of the present invention is shown below. Figure 4 As shown, the device includes:

[0134] The acquisition module 410 is used to acquire the historical sample set, real-time vital sign dataset and real-time prediction results of the identification and early warning model. The historical sample set includes multiple historical prediction results and the historical vital sign dataset corresponding to each historical prediction result.

[0135] The calculation module 420 is used to calculate the confidence level of the real-time prediction results based on the real-time vital signs dataset and the historical sample set;

[0136] The generation module 430 is used to determine the importance of each vital sign data in the historical vital sign dataset based on the historical sample set when the confidence of the real-time prediction result is less than or equal to a preset confidence threshold, and to generate the vital sign weight of the vital sign data based on the importance.

[0137] The correction module 440 is used to correct the vital sign data in the real-time vital sign dataset according to the vital sign weight, generate a corrected vital sign dataset, and input the corrected vital sign dataset into the identification and early warning model to generate the corrected risk probability.

[0138] The output module 450 is used to generate a high-risk correction prediction result and output an alarm signal when the corrected risk probability is greater than a preset probability threshold, and to generate a low-risk correction prediction result and output a silent signal when the corrected risk probability is less than or equal to the preset probability threshold.

[0139] The alarm accuracy improvement device provided in this embodiment can acquire historical sample sets, real-time vital sign datasets, and real-time prediction results of the identification and early warning model through an acquisition module. It calculates the reliability of the real-time prediction results through a calculation module, and generates vital sign weights through a generation module when the reliability of the real-time prediction results is less than or equal to a preset reliability threshold. Subsequently, a correction module corrects the real-time vital sign dataset to generate a corrected risk probability. Finally, an output module generates a corrected prediction result and determines whether to trigger an alarm based on the type of the corrected prediction result. Compared with existing technologies, this device avoids significantly increasing computational and data acquisition costs and makes the corrected risk probability more sensitive through vital sign weights. This allows the prediction results to more accurately reflect the patient's actual condition, thereby improving the alarm accuracy of the EHR system and reducing the probability of false alarms and missed alarms.

[0140] Based on the above embodiments, the computing module includes:

[0141] The first calculation unit is used to calculate the consistency metric value of each historical vital sign dataset in the high-risk historical sample subset, and sort the consistency metric values ​​of each historical vital sign dataset in ascending order to generate a high-risk historical consistency list.

[0142] The second calculation unit is used to calculate the consistency metric value of each historical vital sign dataset in the low-risk historical sample subset, and sort the consistency metric values ​​of each historical vital sign dataset in ascending order to generate a low-risk historical consistency list.

[0143] The first credibility determination unit is used to, when the real-time prediction result is a high-risk real-time prediction result, take the real-time vital signs dataset as a new vital signs dataset in the high-risk historical sample subset, calculate the consistency metric value of the new vital signs dataset, and determine the credibility of the real-time prediction result based on the sorting position of the consistency metric value of the new vital signs dataset in the high-risk historical consistency list.

[0144] The second credibility determination unit is used to, when the real-time prediction result is a low-risk real-time prediction result, take the real-time vital signs dataset as a new vital signs dataset in the low-risk historical sample subset, calculate the consistency metric value of the new vital signs dataset, and determine the credibility of the real-time prediction result based on the sorting position of the consistency metric value of the new vital signs dataset in the low-risk historical consistency list.

[0145] Based on the above embodiments, the device includes:

[0146] The first adjustment module is used to determine the normal range of the historical vital signs dataset based on the historical sample set; obtain the direction and rate of change of the real-time vital signs dataset within a preset time period; if the real-time vital signs dataset changes towards the normal range, the preset confidence threshold is lowered and adjusted, and the step size of the lowering adjustment of the preset confidence threshold is inversely proportional to the rate of change; if the real-time vital signs dataset changes away from the normal range, the preset confidence threshold is raised and adjusted, and the step size of the raising adjustment of the preset confidence threshold is directly proportional to the rate of change.

[0147] Based on the above embodiments, the device includes:

[0148] The second adjustment module is used to obtain the false alarm rate and the missed alarm rate of the EHR system, increase the preset confidence threshold when the false alarm rate increases, and decrease the preset confidence threshold when the missed alarm rate increases.

[0149] Based on the above embodiments, the device includes:

[0150] The third adjustment module is used to determine the normal range of the historical vital sign dataset based on the historical sample set; obtain the object affiliation of the real-time vital sign dataset and the direction of change of the real-time vital sign dataset within a preset time period; if the real-time vital sign dataset changes towards the normal range, the vital sign weight is reduced, and the reduction adjustment step size when the real-time vital sign dataset belongs to a high-risk object is smaller than the reduction adjustment step size when the real-time vital sign dataset belongs to a normal object; if the real-time vital sign dataset changes away from the normal range, the vital sign weight is increased, and the increase adjustment step size when the real-time vital sign dataset belongs to a high-risk object is larger than the increase adjustment step size when the real-time vital sign dataset belongs to a normal object.

[0151] Based on the above embodiments, the device further includes:

[0152] The real-time prediction result judgment module is used to determine the type of real-time prediction result when the credibility of the real-time prediction result is greater than the preset credibility threshold. If the real-time prediction result is a high-risk real-time prediction result, an alarm signal is output; if the real-time prediction result is a low-risk real-time prediction result, a silence signal is output.

[0153] The alarm accuracy improvement device provided in this embodiment of the invention can execute the alarm accuracy improvement method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0154] Example 5

[0155] Figure 5 This is a schematic diagram of the structure of a device provided in Embodiment 5 of the present invention. Figure 5 A block diagram of an exemplary device 12 suitable for implementing embodiments of the present invention is shown. Figure 5 The device 12 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0156] like Figure 5 As shown, device 12 is represented as a general-purpose computing device. Components of device 12 may include, but are not limited to: one or more processors or processing units 16, system memory 28, and a bus 18 connecting different system components (including system memory 28 and processing unit 16).

[0157] Bus 18 represents one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. For example, these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0158] Device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by device 12, including volatile and non-volatile media, removable and non-removable media.

[0159] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache 32. Device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media (… Figure 5 Not shown; usually referred to as a "hard drive"). Although Figure 5 Not shown, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. System memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0160] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. Program modules 42 typically perform the functions and / or methods described in the embodiments of the present invention.

[0161] Device 12 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), and with one or more devices that enable a user to interact with device 12, and / or with any device that enables device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via input / output (I / O) interface 22. Furthermore, device 12 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 20. As shown, network adapter 20 communicates with other modules of device 12 via bus 18. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0162] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the alarm accuracy improvement method provided in the embodiments of the present invention.

[0163] Example 6

[0164] Embodiment 6 of the present invention also provides a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform any of the alarm accuracy improvement methods provided in the above embodiments.

[0165] The computer storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0166] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0167] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including—but not limited to—wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0168] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as "C" or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0169] Example 7

[0170] This invention provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the alarm accuracy improvement method provided in the above embodiments.

[0171] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A method for improving alarm accuracy, characterized in that... include: Acquire the historical sample set, real-time vital sign dataset, and real-time prediction results of the identification and early warning model. The historical sample set includes multiple historical prediction results and the historical vital sign dataset corresponding to each historical prediction result. The reliability of real-time prediction results is calculated based on real-time vital signs datasets and historical sample sets. Determine the normal range of historical vital signs dataset based on historical sample sets; The direction and rate of change of the real-time vital signs dataset within a preset time period are obtained. If the real-time vital signs dataset changes towards the normal range, the preset confidence threshold is lowered and adjusted, and the step size of the lowering of the preset confidence threshold is inversely proportional to the rate of change. If the real-time vital signs dataset changes away from the normal range, the preset confidence threshold is raised and adjusted, and the step size of the raising of the preset confidence threshold is directly proportional to the rate of change. When the confidence level of the real-time prediction result is less than or equal to the preset confidence level threshold, the importance of each vital sign data in the historical vital sign dataset is determined based on the historical sample set, and the vital sign weight of the vital sign data is generated based on the importance level. The system obtains the object attribution of the real-time vital signs dataset and the direction of change of the real-time vital signs dataset within a preset time period. If the real-time vital signs dataset changes towards the normal range, the weight of the vital signs is reduced. The reduction adjustment step size when the real-time vital signs dataset belongs to a high-risk object is smaller than the reduction adjustment step size when the real-time vital signs dataset belongs to a normal object. If the real-time vital signs dataset changes away from the normal range, the weight of the vital signs is increased. The increase adjustment step size when the real-time vital signs dataset belongs to a high-risk object is larger than the increase adjustment step size when the real-time vital signs dataset belongs to a normal object. The vital signs data in the real-time vital signs dataset are corrected according to the vital signs weights to generate a corrected vital signs dataset. The corrected vital signs dataset is then input into the identification and early warning model to generate a corrected risk probability. When the corrected risk probability is greater than the preset probability threshold, a high-risk corrected prediction result is generated and an alarm signal is output. When the corrected risk probability is less than or equal to the preset probability threshold, a low-risk corrected prediction result is generated and a silent signal is output.

2. The alarm accuracy improvement method according to claim 1, characterized in that: The real-time prediction results include high-risk real-time prediction results and low-risk real-time prediction results; The historical sample set includes a high-risk historical sample subset and a low-risk historical sample subset; The high-risk historical sample subset includes high-risk historical prediction results and the corresponding historical vital sign datasets; The low-risk historical sample subset includes low-risk historical prediction results and the corresponding historical vital sign datasets; The calculation of the reliability of real-time prediction results based on real-time vital sign datasets and historical sample sets includes: Calculate the consistency metric for each historical vital sign dataset in the high-risk historical sample subset, and sort the consistency metrics of each historical vital sign dataset in ascending order to generate a high-risk historical consistency list. Calculate the consistency metric for each historical vital sign dataset in the low-risk historical sample subset, and sort the consistency metrics of each historical vital sign dataset in ascending order to generate a low-risk historical consistency list. When the real-time prediction result is a high-risk real-time prediction result, the real-time vital signs dataset is used as a new vital signs dataset in the high-risk historical sample subset. The consistency metric of the new vital signs dataset is calculated, and the credibility of the real-time prediction result is determined according to the sorting position of the consistency metric of the new vital signs dataset in the high-risk historical consistency list. When the real-time prediction result is a low-risk real-time prediction result, the real-time vital signs dataset is used as a new vital signs dataset in the low-risk historical sample subset. The consistency metric of the new vital signs dataset is calculated, and the credibility of the real-time prediction result is determined according to the sorting position of the consistency metric of the new vital signs dataset in the low-risk historical consistency list.

3. The alarm accuracy improvement method according to claim 1, characterized in that: Before calculating the reliability of the real-time prediction result based on the real-time vital signs dataset and the historical sample set, the alarm accuracy improvement method further includes: The false alarm rate and missed alarm rate of the EHR system are obtained. When the false alarm rate increases, the preset confidence threshold is increased. When the missed alarm rate increases, the preset confidence threshold is decreased.

4. The alarm accuracy improvement method according to claim 1, characterized in that: After calculating the reliability of the real-time prediction results based on the real-time vital signs dataset and the historical sample set, the alarm accuracy improvement method further includes: When the credibility of the real-time prediction result is greater than the preset credibility threshold, the type of the real-time prediction result is determined. If the real-time prediction result is a high-risk real-time prediction result, an alarm signal is output. If the real-time prediction result is a low-risk real-time prediction result, a silence signal is output.

5. An alarm accuracy improvement device, characterized in that, include: The acquisition module is used to acquire the historical sample set, real-time vital sign dataset, and real-time prediction results of the identification and early warning model. The historical sample set includes multiple historical prediction results and the historical vital sign dataset corresponding to each historical prediction result. The calculation module is used to calculate the reliability of real-time prediction results based on real-time vital signs datasets and historical sample sets; The first adjustment module is used to determine the normal range of the historical vital signs dataset based on the historical sample set; The direction and rate of change of the real-time vital signs dataset within a preset time period are obtained. If the real-time vital signs dataset changes towards the normal range, the preset confidence threshold is lowered and adjusted, and the step size of the lowering of the preset confidence threshold is inversely proportional to the rate of change. If the real-time vital signs dataset changes away from the normal range, the preset confidence threshold is raised and adjusted, and the step size of the raising of the preset confidence threshold is directly proportional to the rate of change. The generation module is used to determine the importance of each vital sign data in the historical vital sign dataset based on the historical sample set when the confidence of the real-time prediction result is less than or equal to a preset confidence threshold, and to generate the vital sign weight of the vital sign data based on the importance. The third adjustment module is used to obtain the object attribution of the real-time vital signs dataset and the direction of change of the real-time vital signs dataset within a preset time period. If the real-time vital signs dataset changes towards the normal range, the weight of the vital signs is reduced. The reduction adjustment step size when the real-time vital signs dataset belongs to a high-risk object is smaller than the reduction adjustment step size when the real-time vital signs dataset belongs to a normal object. If the real-time vital signs dataset changes away from the normal range, the weight of the vital signs is increased. The increase adjustment step size when the real-time vital signs dataset belongs to a high-risk object is larger than the increase adjustment step size when the real-time vital signs dataset belongs to a normal object. The correction module is used to correct the vital sign data in the real-time vital sign dataset according to the vital sign weight, generate a corrected vital sign dataset, and input the corrected vital sign dataset into the identification and early warning model to generate the corrected risk probability. The output module is used to generate a high-risk correction prediction result and output an alarm signal when the corrected risk probability is greater than a preset probability threshold, and to generate a low-risk correction prediction result and output a silent signal when the corrected risk probability is less than or equal to the preset probability threshold.

6. A device, characterized in that, The device includes: One or more processors; Storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the alarm accuracy improvement method as described in any one of claims 1-4.

7. A storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform the alarm accuracy improvement method as described in any one of claims 1-4.

8. A computer program product, comprising a computer program, characterized in that: When the computer program is executed by the processor, it implements the alarm accuracy improvement method as described in any one of claims 1-4.

Images