A data analysis-based intensive care unit data management system and method

Abstract

This invention discloses a data analysis-based intensive care unit (ICU) data management system and method, relating to the field of ICU management technology. The management method includes the following steps: analyzing and integrating changes in patients' clinical status; capturing alarm nodes during changes in patients' clinical status and generating alarm records; effectively judging any alarm record; conducting difference analysis between different alarm records; clustering patients' similar clinical statuses based on the differences in alarm records at each alarm node; extracting the patient's clinical status from any alarm record, effectively evaluating the alarm records, and individually optimizing the alarm thresholds for effective alarms; performing real-time alarm evaluation on patients' real-time clinical data and sending alarm reminders when personalized alarm thresholds are met; effectively reducing false alarms caused by fixed threshold settings, ensuring resources are focused on critical events, and improving clinical response efficiency.

Description

Technical Field

[0001] This invention relates to the field of monitoring and management technology, specifically a critical care data management system and method based on data analysis. Background Technology

[0002] Intensive care units are the core units for treating critically ill patients in modern hospitals. They continuously generate massive amounts of high-dimensional and dynamic physiological monitoring data and treatment information. Effective management and utilization of this data are crucial for real-time assessment of patient status and timely detection of deterioration of condition.

[0003] Currently, intensive care data is generally collected through bedside monitors with fixed threshold alarm systems. However, fixed thresholds cannot adapt to individual patient differences and dynamic changes in their condition. Different patients have significantly different conditions and responses, making it difficult for existing systems to set personalized alarm thresholds. This can easily lead to false alarms or missed alarms, causing "alarm fatigue" among medical staff, seriously interfering with the efficiency of clinical decision-making, and potentially overlooking truly critical events. Summary of the Invention

[0004] The purpose of this invention is to provide a critical care data management system and method based on data analysis to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a data analysis-based method for managing intensive care unit (ICU) data, the method comprising the following steps:

[0006] Step S1: Collect historical clinical data of patients in the intensive care unit, analyze and integrate changes in patients' clinical status; capture alarm nodes in the process of changes in patients' clinical status and generate corresponding alarm records;

[0007] Step S2: Collect clinical intervention data for any alarm record and make an effective judgment on the alarm record; based on the different effective judgment results between different alarm records, conduct a difference analysis between different alarm records;

[0008] Step S3: Compare and analyze the changes in clinical status of different patients, and cluster patients with similar clinical statuses based on the differences in alarm records at each alarm node;

[0009] Step S4: Extract the patient's clinical status from any alarm record, and evaluate the effectiveness of the alarm record based on the comparison with similar clinical statuses; based on the effectiveness judgment results of the alarm record, optimize the alarm threshold of the effective alarm.

[0010] Step S5: Perform real-time alert assessment on the patient's real-time clinical data, extract the patient's individual alert thresholds, and send alert reminders when the individual alert thresholds are met.

[0011] Furthermore, step S1 includes the following steps:

[0012] Step S11: Collect the patient's clinical data through the monitoring equipment at the patient's bedside, obtain several preset monitoring indicators in the monitoring equipment, divide the collected clinical data according to each monitoring indicator, and obtain the clinical dataset for each monitoring indicator; the preset monitoring indicators in the monitoring equipment include electrocardiogram monitoring such as heart rate and heart rhythm, blood pressure monitoring such as blood pressure fluctuation, and respiratory monitoring such as respiratory rate, etc.

[0013] Step S12: Sort the clinical datasets of each monitoring indicator according to the order of collection time points and place them in a two-dimensional Cartesian coordinate system to present the changing trend of each monitoring indicator over time; arbitrarily select a collection time point and integrate the clinical data corresponding to each monitoring indicator at the selected collection time point to obtain the user's clinical status data group at the selected collection time point; the temporary status data group is a summary of the real-time values ​​of the patient's heart rate, blood pressure, respiratory rate and other monitoring indicators at any time point. By merging all monitoring indicators, the user's clinical status can be reflected;

[0014] Step S13: Whenever a patient triggers an alarm at a certain time point, the certain time point is set as an alarm node, and the clinical status data group at the certain time point is extracted. At the same time, the patient's other clinical information is collected to generate an alarm record for the patient at the certain time point.

[0015] Furthermore, step S2 includes the following steps:

[0016] Step S21: Whenever an alarm record is generated, the alarm response behavior of medical staff is collected, and features are extracted from the alarm response behavior of medical staff to obtain several behavioral features of medical staff; by recording the response behavior of medical staff in text form, and extracting features from the text, if the response behavior contains features such as "adjust", "handle", and "take action", it indicates that the alarm is valid; if it contains features such as "ignore" or "invalid", it indicates that the alarm is invalid.

[0017] Step S22: A database of invalid features is pre-constructed. The database contains several invalid features. Each of the several behavioral features is compared with the database of invalid features. If a behavioral feature is the same as an invalid feature, the generated alarm record is marked as invalid. If no behavioral feature is different from any invalid feature, the generated alarm record is marked as valid. The database of invalid features stores several behavioral terms that can reflect the invalidity of the alarm, such as ignore, invalid, and normal.

[0018] Step S23: Acquire several alarm records during the patient's clinical status change process, randomly select one alarm record from the several alarm records, extract the clinical status data group of the selected alarm record; select the clinical data of the i-th monitoring indicator from the clinical status data group. If the selected alarm record has an invalid mark, set the clinical data of the i-th monitoring indicator as abnormal data; if the selected alarm record has a valid mark, set it as normal data.

[0019] Step S24: Divide and summarize the clinical data of the i-th monitoring indicator in the plurality of alarm records according to the labels to obtain the normal data range of the i-th monitoring indicator as (d1) min ,d1 max ) and the range of abnormal data (d2) min ,d2 max If there is a common range between two data ranges, the i-th monitoring indicator is set as the first feature marker; if there is no common range between the two data ranges, the preset alarm threshold (d) of the i-th monitoring indicator is obtained. i ) th If d1 max <(d) i ) th <d2 min Or d2 max <(d) i ) th <d1 min If the first feature flag indicates that the monitoring indicator may have false alarms according to the preset alarm threshold, that is, the alarm threshold setting is unreasonable; if the second feature flag indicates that the monitoring indicator alarms normally according to the preset alarm threshold, then the alarm threshold setting is reasonable and effective.

[0020] Step S25: Acquire all alarm records for different patients and mark each monitoring indicator with features. If the i-th monitoring indicator has a first abnormal marker in the alarm records of different patients, then the i-th monitoring indicator is set as a personalized monitoring indicator. If the i-th monitoring indicator has a second feature marker in the alarm records of different patients, then the i-th monitoring indicator is set as a routine monitoring indicator. Perform a comprehensive analysis on the feedback of the same monitoring indicator in the alarm records of different patients. If false alarms occur in different patients, it indicates that there are individual differences in the monitoring indicator, and personalized adjustments need to be made for different patients. If no false alarms occur, it means that the preset threshold can continue to be used.

[0021] Furthermore, step S3 includes the following steps:

[0022] Step S31: Randomly select all alarm records of a patient, and randomly select one target alarm record from all alarm records. Randomly select the i-th monitoring indicator from the selected target alarm record. If the i-th monitoring indicator is a personalized monitoring indicator, then set the i-th monitoring indicator as the target monitoring indicator.

[0023] Step S32: Randomly select one comparison alarm record from the remaining patients. If the i-th monitoring indicator is a routine monitoring indicator in the selected comparison alarm record, then set the i-th monitoring indicator as a difference indicator. If the i-th monitoring indicator is a personalized monitoring indicator in the selected comparison alarm record, then obtain the normal data range and abnormal data range of the i-th monitoring indicator in the two alarm records respectively. Set the common range of the selected target alarm record as (d3, d4) and the common range of the selected comparison alarm record as (d5, d6), and obtain the size Δd1 of the common range of the i-th monitoring indicator in the selected target alarm record. i =d4-d3 and the common range size Δd2 in the selected comparison alarm records. i =d6-d5, the deviation between the two common ranges is calculated as p. i =|Δd1 i -Δd2 i | / Max(d6-d3,d4-d5), where Max() is the function to find the maximum value; a preset deviation threshold p is used. th If p i <p th If so, then the i-th monitoring indicator will be set as a similar indicator;

[0024] Step S33: Compare the i-th monitoring indicator in each alarm record for any two patients, and count the number of times the i-th monitoring indicator is the same type of indicator between any two alarm records, which is m. i Let m be the total number of comparisons for the i-th monitoring indicator. totalThe percentage of times the i-th monitoring indicator is the same type of indicator is calculated to be η. i =m i / m total Preset a threshold value η for the percentage of occurrences. th If η i >η th If so, then the i-th monitoring indicator will be set as a similar indicator;

[0025] Step S34: Count the number of similar indicators between the two patients as r, and set the total number of monitoring indicators as r. ’ The proportion of similar indicators is obtained as δ = r / r ’ Preset a quantity percentage threshold δ th If δ≥δ th Then, the changes in the clinical status of the two patients are set as the same type of clinical status, and all the same type of clinical status of the remaining patients are divided into clusters.

[0026] Furthermore, step S4 includes the following steps:

[0027] Step S41: Randomly select an alarm record from the alarm node in the process of a patient's clinical status change. Randomly select the i-th monitoring indicator from the selected alarm record to obtain the clinical data d of the selected alarm record for the i-th monitoring indicator. i ;

[0028] Step S42: When the i-th monitoring indicator is a regular monitoring indicator, obtain the preset alarm threshold (d) for the i-th monitoring indicator. i ) th The degree of abnormality of the i-th monitoring indicator is obtained as Y. i =[d i -(d i ) th ] / (d i ) th When the i-th monitoring indicator is a personalized monitoring indicator, if the i-th monitoring indicator is the first feature marker in the selected alarm records, then obtain the abnormal data range of the i-th monitoring indicator in the alarm nodes of all patients in the same clinical state, and select the abnormal data with the smallest value as the updated alarm threshold (d) for the i-th monitoring indicator. i ) ’ th The degree of abnormality of the i-th monitoring indicator is obtained as Y. i =[d i -(d i ) ’ th ] / (d i ) ’ thIf the i-th monitoring indicator is the second feature marker in the selected alarm records, then the feature marker of the i-th monitoring indicator in all alarm nodes of the patient is obtained. If all of them are the second feature markers, then the abnormality degree of the i-th monitoring indicator is Y. i =[d i -(d i ) th ] / (d i ) th If a first feature marker exists, then the degree of abnormality of the i-th monitoring indicator is obtained as Y. i =[d i -(d i ) ’ th ] / (d i ) ’ th ;

[0029] Step S43: Obtain the degree of abnormality for each monitoring indicator in the selected alarm records, and set the effective alarm value E for the i-th monitoring indicator. i If Y i If ≤0, then E i =0, if Y i If E > 0, then E i =1-Y i According to the formula:

[0030] ;

[0031] The effective evaluation value G of the selected alarm record is calculated;

[0032] Step S44: Identify the markers of the selected alarm records. If the selected alarm record is a valid marker, set the valid evaluation value as the feature evaluation value. If the selected alarm record is an invalid marker, set the valid evaluation value as the abnormal evaluation value. Obtain the valid evaluation values ​​for each alarm record of all patients in the same clinical state. Divide all valid evaluation values ​​to generate feature evaluation value ranges and abnormal evaluation value ranges respectively. If the two evaluation value ranges overlap, select the minimum abnormal evaluation value as the valid alarm threshold for the same clinical state of a certain patient. If the two evaluation value ranges do not overlap, select the maximum feature evaluation value as the valid alarm threshold for the same clinical state of a certain patient.

[0033] Furthermore, step S5 includes the following steps:

[0034] Step S51: Acquire clinical status data sets of any patient in real time, obtain clinical data for each monitoring indicator, and set the real-time clinical data of the i-th monitoring indicator as (d i ) nowThe alarm threshold for the patient on the i-th monitoring indicator is obtained as (d i ) ’’ th The real-time anomaly level of the i-th monitoring indicator is obtained as (Y). i ) now =[(d i ) now -(d i ) ’’ th ] / (d i ) ’’ th ;

[0035] Step S52: Obtain the real-time anomaly level of each monitoring indicator and set a real-time valid value (E). i ) now If (Y) i ) now ≤0, then (E i ) now =0, if (Y) i ) now >0, then (E i ) now =1-Y i The patient's real-time assessment value G is obtained by summing the values. now Set the effective alarm threshold for patients to G. th If G now ≥G th Then, an alert will be sent to remind the patient of their current clinical status.

[0036] To better implement the above methods, an intensive care data management system is also proposed. The management system includes a historical alarm collection module, an alarm difference analysis module, a clinical similarity clustering module, an alarm evaluation and optimization module, and a real-time anomaly judgment module.

[0037] The historical alarm acquisition module is used to collect historical clinical data of patients in intensive care units, analyze and integrate changes in patients' clinical status; capture alarm nodes in the process of changes in patients' clinical status, and generate corresponding alarm records.

[0038] The alarm difference analysis module is used to collect clinical intervention information for any alarm record and make effective judgments on the alarm records; based on the different effective judgment results between different alarm records, it conducts difference analysis between different alarm records;

[0039] The clinical similarity clustering module is used to compare and analyze the changes in the clinical status of different patients. Based on the differences in alarm records at each alarm node, patients of the same type are clustered.

[0040] The alarm assessment and optimization module is used to extract the patient's clinical status from any alarm record, and to perform an effective alarm assessment on the alarm record based on comparison with similar records; based on the effective judgment results of the alarm record, the alarm threshold of the effective alarm is individually optimized.

[0041] The real-time anomaly detection module is used to perform real-time alarm assessment on the patient's real-time clinical data, extract the patient's individual alarm thresholds, and send alarm reminders when the individual alarm thresholds are met.

[0042] Furthermore, the historical alarm acquisition module includes a clinical status integration unit and an alarm record generation unit;

[0043] The clinical status integration unit is used to collect historical clinical data of patients in the intensive care unit, analyze changes in patients' clinical status and integrate them; the alarm record generation unit is used to capture alarm nodes in the process of changes in patients' clinical status and generate corresponding alarm records.

[0044] Furthermore, the alarm difference analysis module includes an alarm validity identification unit and an alarm difference analysis unit;

[0045] The alarm effective identification unit is used to collect clinical intervention information for any alarm record and to make an effective judgment on the alarm record; the alarm difference analysis unit is used to conduct difference analysis between different alarm records based on the different effective judgment results between different alarm records.

[0046] Furthermore, the alarm assessment and optimization module includes an effective alarm assessment unit and a personalized threshold optimization unit;

[0047] The effective alarm assessment unit is used to extract the patient's clinical status from any alarm record and assess the effectiveness of the alarm record based on comparison with similar records; the personalized threshold optimization unit is used to optimize the alarm threshold of the effective alarm based on the effective judgment results of the alarm record.

[0048] Compared with the prior art, the beneficial effects of the present invention are:

[0049] 1. This invention establishes an effective alarm judgment mechanism and effective alarm evaluation, dynamically distinguishes between critical events and invalid alarms, reduces false alarms caused by fixed threshold settings, effectively alleviates "alarm fatigue" of medical staff, ensures that resources are focused on real critical events, and improves clinical response efficiency;

[0050] 2. This invention uses clinical status clustering and personalized threshold optimization to customize monitoring indicator thresholds for different patients, which can adapt to individual differences and dynamic changes in the condition of patients, avoid missed reports caused by uniform thresholds, and improve monitoring accuracy.

[0051] 3. This invention integrates multi-dimensional physiological indicator data and uses a quantitative model to comprehensively assess the patient's condition, capture the trend of disease deterioration in real time, and assist medical staff in making rapid decisions; thereby improving the overall diagnostic and treatment efficiency and resource utilization rate of intensive care units. Attached Figure Description

[0052] Figure 1 This is a schematic diagram illustrating the steps of a data analysis-based intensive care data management method.

[0053] Figure 2 This is a schematic diagram of the structure of a data analysis-based intensive care data management system;

[0054] Figure 3 This is a flowchart. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Example: Figures 1 to 3 As shown, this invention provides a data analysis-based method for managing intensive care unit (ICU) data. The method includes the following steps:

[0057] Step S1: Collect historical clinical data of patients in the intensive care unit, analyze and integrate changes in patients' clinical status; capture alarm nodes in the process of changes in patients' clinical status and generate corresponding alarm records;

[0058] Step S1 includes the following steps:

[0059] Step S11: Collect the patient's clinical data through the monitoring equipment at the patient's bedside, obtain several preset monitoring indicators in the monitoring equipment, divide the collected clinical data according to each monitoring indicator, and obtain the clinical dataset for each monitoring indicator.

[0060] Step S12: Sort the clinical datasets of each monitoring indicator according to the order of collection time points and place them in a two-dimensional Cartesian coordinate system to present the changing trend of each monitoring indicator over time; arbitrarily select a collection time point and integrate the clinical data corresponding to each monitoring indicator at the selected collection time point to obtain the user's clinical status data group at the selected collection time point.

[0061] Step S13: Whenever a patient triggers an alarm at a certain time point, the certain time point is set as an alarm node, and the clinical status data group at the certain time point is extracted. At the same time, the patient's other clinical information is collected to generate an alarm record for the patient at the certain time point.

[0062] Step S2: Collect clinical intervention data for any alarm record and make an effective judgment on the alarm record; based on the different effective judgment results between different alarm records, conduct a difference analysis between different alarm records;

[0063] Step S2 includes the following steps:

[0064] Step S21: Whenever an alarm record is generated, the alarm response behavior of medical staff is collected, and the alarm response behavior of medical staff is feature extracted to obtain several behavioral features of medical staff.

[0065] Step S22: Pre-build an invalid feature database containing several invalid features. Compare each of the several behavioral features with the invalid feature database. If a behavioral feature is the same as an invalid feature, mark the generated alarm record as invalid. If any behavioral feature is different from any invalid feature, mark the generated alarm record as valid.

[0066] Step S23: Acquire several alarm records during the patient's clinical status change process, randomly select one alarm record from the several alarm records, extract the clinical status data group of the selected alarm record; select the clinical data of the i-th monitoring indicator from the clinical status data group. If the selected alarm record has an invalid mark, set the clinical data of the i-th monitoring indicator as abnormal data; if the selected alarm record has a valid mark, set it as normal data.

[0067] Step S24: Divide and summarize the clinical data of the i-th monitoring indicator in the plurality of alarm records according to the labels to obtain the normal data range of the i-th monitoring indicator as (d1) min ,d1 max ) and the range of abnormal data (d2) min ,d2 max If there is a common range between two data ranges, the i-th monitoring indicator is set as the first feature marker; if there is no common range between the two data ranges, the preset alarm threshold (d) of the i-th monitoring indicator is obtained. i ) th If d1 max <(d) i ) th <d2 min Or d2max <(d) i ) th <d1 min Then the i-th monitoring indicator is set as the second feature label;

[0068] Example 1: Ten heart rate alarm records of a patient were obtained. The normal values ​​were 105, 110, 118, 103, and 120, and the abnormal values ​​were 108, 122, 107, and 109. That is, the normal data range is (103, 120), and the abnormal data is (107, 122). Since there is a common range (107, 120), the heart rate is marked with the first feature.

[0069] Step S25: Acquire all alarm records for different patients and mark each monitoring indicator with features. If the i-th monitoring indicator has a first abnormal mark in the alarm records of different patients, then the i-th monitoring indicator is set as a personalized monitoring indicator. If the i-th monitoring indicator has a second feature mark in the alarm records of different patients, then the i-th monitoring indicator is set as a routine monitoring indicator.

[0070] Step S3: Compare and analyze the changes in clinical status of different patients, and cluster patients with similar clinical statuses based on the differences in alarm records at each alarm node;

[0071] Step S3 includes the following steps:

[0072] Step S31: Randomly select all alarm records of a patient, and randomly select one target alarm record from all alarm records. Randomly select the i-th monitoring indicator from the selected target alarm record. If the i-th monitoring indicator is a personalized monitoring indicator, then set the i-th monitoring indicator as the target monitoring indicator.

[0073] Step S32: Randomly select one comparison alarm record from the remaining patients. If the i-th monitoring indicator is a routine monitoring indicator in the selected comparison alarm record, then set the i-th monitoring indicator as a difference indicator. If the i-th monitoring indicator is a personalized monitoring indicator in the selected comparison alarm record, then obtain the normal data range and abnormal data range of the i-th monitoring indicator in the two alarm records respectively. Set the common range of the selected target alarm record as (d3, d4) and the common range of the selected comparison alarm record as (d5, d6), and obtain the size Δd1 of the common range of the i-th monitoring indicator in the selected target alarm record. i =d4-d3 and the common range size Δd2 in the selected comparison alarm records. i =d6-d5, the deviation between the two common ranges is calculated as p. i =|Δd1 i -Δd2 i| / Max(d6-d3,d4-d5), where Max() is the function to find the maximum value; a preset deviation threshold p is used. th If p i <p th If so, then the i-th monitoring indicator will be set as a similar indicator;

[0074] Example 2: Set the normal range of the target patient's alarm record to (100, 115) and the abnormal range to (110, 120), resulting in a common range size of 5; set the normal range of the comparison patient's alarm record to (98, 118) and the abnormal range to (112, 120), resulting in a common range size of 6, and obtain the deviation degree p=1 / Max(8,3)=1 / 8=0.125;

[0075] Step S33: Compare the i-th monitoring indicator in each alarm record for any two patients, and count the number of times the i-th monitoring indicator is the same type of indicator between any two alarm records, which is m. i Let m be the total number of comparisons for the i-th monitoring indicator. total The percentage of times the i-th monitoring indicator is the same type of indicator is calculated to be η. i =m i / m total Preset a threshold value η for the percentage of occurrences. th If η i >η th If so, then the i-th monitoring indicator will be set as a similar indicator;

[0076] Step S34: Count the number of similar indicators between the two patients as r, and set the total number of monitoring indicators as r. ’ The proportion of similar indicators is obtained as δ=r / r ’ Preset a quantity percentage threshold δ th If δ≥δ th Then, the changes in the clinical status of the two patients are set as the same type of clinical status, and all the same type of clinical status of the remaining patients are divided into clusters.

[0077] Step S4: Extract the patient's clinical status from any alarm record, and evaluate the effectiveness of the alarm record based on the comparison with similar clinical statuses; based on the effectiveness judgment results of the alarm record, optimize the alarm threshold of the effective alarm.

[0078] Step S4 includes the following steps:

[0079] Step S41: Randomly select an alarm record from the alarm node in the process of a patient's clinical status change. Randomly select the i-th monitoring indicator from the selected alarm record to obtain the clinical data d of the selected alarm record for the i-th monitoring indicator.i ;

[0080] Step S42: When the i-th monitoring indicator is a regular monitoring indicator, obtain the preset alarm threshold (d) for the i-th monitoring indicator. i ) th The degree of abnormality of the i-th monitoring indicator is obtained as Y. i =[d i -(d i ) th ] / (d i ) th When the i-th monitoring indicator is a personalized monitoring indicator, if the i-th monitoring indicator is the first feature marker in the selected alarm records, then obtain the abnormal data range of the i-th monitoring indicator in the alarm nodes of all patients in the same clinical state, and select the abnormal data with the smallest value as the updated alarm threshold (d) for the i-th monitoring indicator. i ) ’ th The degree of abnormality of the i-th monitoring indicator is obtained as Y. i =[d i -(d i ) ’ th ] / (d i ) ’ th If the i-th monitoring indicator is the second feature marker in the selected alarm records, then the feature marker of the i-th monitoring indicator in all alarm nodes of the patient is obtained. If all of them are the second feature markers, then the abnormality degree of the i-th monitoring indicator is Y. i =[d i -(d i ) th ] / (d i ) th If a first feature marker exists, then the degree of abnormality of the i-th monitoring indicator is obtained as Y. i =[d i -(d i ) ’ th ] / (d i ) ’ th ;

[0081] Step S43: Obtain the degree of abnormality for each monitoring indicator in the selected alarm records, and set the effective alarm value E for the i-th monitoring indicator. i If Y i If ≤0, then E i =0, if Y i If E > 0, then E i =1-Y i According to the formula:

[0082] ;

[0083] The effective evaluation value G of the selected alarm record is calculated;

[0084] Step S44: Identify the markers of the selected alarm records. If the selected alarm record is a valid marker, set the valid evaluation value as the feature evaluation value. If the selected alarm record is an invalid marker, set the valid evaluation value as the abnormal evaluation value. Obtain the valid evaluation values ​​for each alarm record of all patients in the same clinical state. Divide all valid evaluation values ​​to generate feature evaluation value ranges and abnormal evaluation value ranges respectively. If the two evaluation value ranges overlap, select the minimum abnormal evaluation value as the effective alarm threshold for the same clinical state of the patient. If the two evaluation value ranges do not overlap, select the maximum feature evaluation value as the effective alarm threshold for the same clinical state of the patient.

[0085] Step S5: Perform real-time alarm assessment on the patient's real-time clinical data, extract the patient's personalized alarm threshold, and send alarm reminders when the personalized alarm threshold is met;

[0086] Step S5 includes the following steps:

[0087] Step S51: Acquire clinical status data sets of any patient in real time, obtain clinical data for each monitoring indicator, and set the real-time clinical data of the i-th monitoring indicator as (d i ) now The alarm threshold for the patient on the i-th monitoring indicator is obtained as (d i ) ’’ th The real-time anomaly level of the i-th monitoring indicator is obtained as (Y). i ) now =[(d i ) now -(d i ) ’’ th ] / (d i ) ’’ th ;

[0088] Step S52: Obtain the real-time anomaly level of each monitoring indicator and set a real-time valid value (E). i ) now If (Y) i ) now ≤0, then (E i ) now =0, if (Y) i ) now >0, then (E i )now =1-Y i The patient's real-time assessment value G is obtained by summing the values. now Set the effective alarm threshold for patients to G. th If G now ≥G th Then, an alert will be sent to remind the patient of their current clinical status.

[0089] An intensive care unit data management system, comprising a historical alarm acquisition module, an alarm difference analysis module, a clinical similarity clustering module, an alarm evaluation and optimization module, and a real-time anomaly judgment module;

[0090] The historical alarm acquisition module is used to collect historical clinical data of patients in intensive care units, analyze and integrate changes in patients' clinical status; capture alarm nodes in the process of changes in patients' clinical status, and generate corresponding alarm records.

[0091] The alarm difference analysis module is used to collect clinical intervention information for any alarm record and make effective judgments on the alarm records; based on the different effective judgment results between different alarm records, it conducts difference analysis between different alarm records;

[0092] The clinical similarity clustering module is used to compare and analyze the changes in the clinical status of different patients. Based on the differences in alarm records at each alarm node, patients of the same type are clustered.

[0093] The alarm assessment and optimization module is used to extract the patient's clinical status from any alarm record, and to perform an effective alarm assessment on the alarm record based on comparison with similar records; based on the effective judgment results of the alarm record, the alarm threshold of the effective alarm is individually optimized.

[0094] The real-time anomaly detection module is used to perform real-time alarm assessment on the patient's real-time clinical data, extract the patient's individual alarm thresholds, and send alarm reminders when the individual alarm thresholds are met.

[0095] The historical alarm acquisition module includes a clinical status integration unit and an alarm record generation unit;

[0096] The clinical status integration unit is used to collect historical clinical data of patients in the intensive care unit, analyze changes in patients' clinical status and integrate them; the alarm record generation unit is used to capture alarm nodes in the process of changes in patients' clinical status and generate corresponding alarm records.

[0097] The alarm difference analysis module includes an alarm effective identification unit and an alarm difference analysis unit.

[0098] The alarm effective identification unit is used to collect clinical intervention information for any alarm record and to make an effective judgment on the alarm record; the alarm difference analysis unit is used to conduct difference analysis between different alarm records based on the different effective judgment results between different alarm records.

[0099] The alarm assessment and optimization module includes an effective alarm assessment unit and a personalized threshold optimization unit.

[0100] The effective alarm assessment unit is used to extract the patient's clinical status from any alarm record and assess the effectiveness of the alarm record based on comparison with similar records; the personalized threshold optimization unit is used to optimize the alarm threshold of the effective alarm based on the effective judgment results of the alarm record.

[0101] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A data analysis-based method for managing critical care data, characterized in that: The management method includes the following steps: Step S1: Collect historical clinical data of patients in the intensive care unit, analyze and integrate changes in patients' clinical status; capture alarm nodes in the process of changes in patients' clinical status and generate corresponding alarm records; Step S2: Collect clinical intervention data for any alarm record and make an effective judgment on the alarm record; based on the different effective judgment results between different alarm records, conduct a difference analysis between different alarm records; Step S3: Compare and analyze the changes in clinical status of different patients, and cluster patients with similar clinical statuses based on the differences in alarm records at each alarm node; Step S4: Extract the patient's clinical status from any alarm record, and evaluate the effectiveness of the alarm record based on the comparison with similar clinical statuses; optimize the alarm threshold for effective alarms based on the effective judgment results of the alarm record. Step S5: Perform real-time alarm assessment on the patient's real-time clinical data, extract the patient's personalized alarm threshold, and send alarm reminders when the personalized alarm threshold is met; Step S2 includes the following steps: Step S21: Whenever an alarm record is generated, the alarm response behavior of medical staff is collected, and the alarm response behavior of medical staff is feature extracted to obtain several behavioral features of medical staff. Step S22: Pre-build an invalid feature database containing several invalid features. Compare each of the several behavioral features with the invalid feature database. If a behavioral feature is the same as an invalid feature, mark the generated alarm record as invalid. If any behavioral feature is different from any invalid feature, mark the generated alarm record as valid. Step S23: Acquire several alarm records during the patient's clinical status change process, randomly select one alarm record from the several alarm records, extract the clinical status data group of the selected alarm record; select the clinical data of the i-th monitoring indicator from the clinical status data group. If the selected alarm record has an invalid mark, set the clinical data of the i-th monitoring indicator as abnormal data; if the selected alarm record has a valid mark, set it as normal data. Step S24: Divide and summarize the clinical data of the i-th monitoring indicator in the plurality of alarm records according to the labels to obtain the normal data range of the i-th monitoring indicator as (d1) min ,d1 max ) and the range of abnormal data (d2) min ,d2 max If there is a common range between two data ranges, the i-th monitoring indicator is set as the first feature marker; if there is no common range between the two data ranges, the preset alarm threshold (d) of the i-th monitoring indicator is obtained. i ) th If d1 max <(d) i ) th <d2 min Or d2 max <(d) i ) th <d1 min Then the i-th monitoring indicator is set as the second feature label; Step S25: Acquire all alarm records for different patients and mark each monitoring indicator with features. If the i-th monitoring indicator has a first abnormal mark in the alarm records of different patients, then the i-th monitoring indicator is set as a personalized monitoring indicator. If the i-th monitoring indicator has a second feature mark in the alarm records of different patients, then the i-th monitoring indicator is set as a routine monitoring indicator.

2. The critical care data management method based on data analysis according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Collect the patient's clinical data through the monitoring equipment at the patient's bedside, obtain several preset monitoring indicators in the monitoring equipment, divide the collected clinical data according to each monitoring indicator, and obtain the clinical dataset for each monitoring indicator. Step S12: Sort the clinical datasets of each monitoring indicator according to the order of collection time points and place them in a two-dimensional Cartesian coordinate system to present the changing trend of each monitoring indicator over time; arbitrarily select a collection time point and integrate the clinical data corresponding to each monitoring indicator at the selected collection time point to obtain the user's clinical status data group at the selected collection time point. Step S13: Whenever a patient triggers an alarm at a certain time point, the certain time point is set as an alarm node, and the clinical status data group at the certain time point is extracted. At the same time, the patient's other clinical information is collected to generate an alarm record for the patient at the certain time point.

3. The critical care data management method based on data analysis according to claim 2, characterized in that: Step S3 includes the following steps: Step S31: Randomly select all alarm records of a patient, and randomly select one target alarm record from all alarm records. Randomly select the i-th monitoring indicator from the selected target alarm record. If the i-th monitoring indicator is a personalized monitoring indicator, then set the i-th monitoring indicator as the target monitoring indicator. Step S32: Randomly select one comparison alarm record from the remaining patients. If the i-th monitoring indicator is a routine monitoring indicator in the selected comparison alarm record, then set the i-th monitoring indicator as a difference indicator. If the i-th monitoring indicator is a personalized monitoring indicator in the selected comparison alarm record, then obtain the normal data range and abnormal data range of the i-th monitoring indicator in the two alarm records respectively. Set the common range of the selected target alarm record as (d3, d4) and the common range of the selected comparison alarm record as (d5, d6), and obtain the size Δd1 of the common range of the i-th monitoring indicator in the selected target alarm record. i =d4-d3 and the common range size Δd2 in the selected comparison alarm records. i =d6-d5, the deviation between the two common ranges is calculated as p. i =|Δd1 i -Δd2 i | / Max(d6-d3,d4-d5), where Max() is the function to find the maximum value; a preset deviation threshold p is used. th If p i <p th If so, then the i-th monitoring indicator will be set as a similar indicator; Step S33: Compare the i-th monitoring indicator in each alarm record for any two patients, and count the number of times the i-th monitoring indicator is the same type of indicator between any two alarm records, which is m. i Let m be the total number of comparisons for the i-th monitoring indicator. total The percentage of times the i-th monitoring indicator is the same type of indicator is calculated to be η. i =m i / m total Preset a threshold value η for the percentage of occurrences. th If η i >η th If so, then the i-th monitoring indicator will be set as a similar indicator; Step S34: Count the number of similar indicators between the two patients as r, and set the total number of monitoring indicators as r. ’ The proportion of similar indicators is obtained as δ=r / r ’ Preset a quantity percentage threshold δ th If δ≥δ th Then, the changes in the clinical status of the two patients are set as the same type of clinical status, and all the same type of clinical status of the remaining patients are divided into clusters.

4. The critical care data management method based on data analysis according to claim 3, characterized in that: Step S4 includes the following steps: Step S41: Randomly select an alarm record from the alarm node in the process of a patient's clinical status change. Randomly select the i-th monitoring indicator from the selected alarm record to obtain the clinical data d of the selected alarm record for the i-th monitoring indicator. i ; Step S42: When the i-th monitoring indicator is a regular monitoring indicator, obtain the preset alarm threshold (d) for the i-th monitoring indicator. i ) th The degree of abnormality of the i-th monitoring indicator is obtained as Y. i =[d i -(d i ) th ] / (d i ) th When the i-th monitoring indicator is a personalized monitoring indicator, if the i-th monitoring indicator is the first feature marker in the selected alarm records, then obtain the abnormal data range of the i-th monitoring indicator in the alarm nodes of all patients in the same clinical state, and select the abnormal data with the smallest value as the updated alarm threshold (d) for the i-th monitoring indicator. i ) ’ th The degree of abnormality of the i-th monitoring indicator is obtained as Y. i =[d i -(d i ) ’ th ] / (d i ) ’ th If the i-th monitoring indicator is marked with the second feature in the selected alarm records, then the feature label of the i-th monitoring indicator in all alarm nodes of the user is obtained. If all of them are marked with the second feature, then the abnormality degree of the i-th monitoring indicator is Y. i =[d i -(d i ) th ] / (d i ) th If a first feature marker exists, then the degree of abnormality of the i-th monitoring indicator is obtained as Y. i =[d i -(d i ) ’ th ] / (d i ) ’ th ; Step S43: Obtain the degree of abnormality for each monitoring indicator in the selected alarm records, and set the effective alarm value E for the i-th monitoring indicator. i If Y i If ≤0, then E i =0, if Y i If E > 0, then E i =1-Y i According to the formula: ； The effective evaluation value G of the selected alarm record is calculated; Step S44: Identify the markers of the selected alarm records. If the selected alarm record is a valid marker, set the valid evaluation value as the feature evaluation value. If the selected alarm record is an invalid marker, set the valid evaluation value as the abnormal evaluation value. Obtain the valid evaluation values ​​for each alarm record of all patients in the same clinical state. Divide all valid evaluation values ​​to generate feature evaluation value ranges and abnormal evaluation value ranges respectively. If the two evaluation value ranges overlap, select the minimum abnormal evaluation value as the effective alarm threshold for the same clinical state of the patient. If the two evaluation value ranges do not overlap, select the maximum feature evaluation value as the effective alarm threshold for the same clinical state of the patient.

5. The critical care data management method based on data analysis according to claim 4, characterized in that: Step S5 includes the following steps: Step S51: Acquire clinical status data sets of any patient in real time, obtain clinical data for each monitoring indicator, and set the real-time clinical data of the i-th monitoring indicator as (d i ) now The alarm threshold for the patient on the i-th monitoring indicator is obtained as (d i ) ’’ th The real-time anomaly level of the i-th monitoring indicator is obtained as (Y). i ) now =[(d i ) now -(d i ) ’’ th ] / (d i ) ’’ th ; Step S52: Obtain the real-time anomaly level of each monitoring indicator and set a real-time valid value (E). i ) now If (Y) i ) now ≤0, then (E i ) now =0, if (Y) i ) now >0, then (E i ) now =1-Y i The patient's real-time assessment value G is obtained by summing the values. now Set the effective alarm threshold for patients to G. th If G now ≥G th Then, an alert will be sent to remind the patient of their current clinical status.

6. An intensive care unit (ICU) data management system, used to execute the ICU data management method based on data analysis according to any one of claims 1-5, characterized in that: The management system includes a historical alarm acquisition module, an alarm difference analysis module, a clinical similarity clustering module, an alarm evaluation and optimization module, and a real-time anomaly judgment module. The historical alarm acquisition module is used to collect historical clinical data of patients in intensive care units, analyze and integrate changes in patients' clinical status; capture alarm nodes in the process of changes in patients' clinical status, and generate corresponding alarm records. The alarm difference analysis module is used to collect clinical intervention information for any alarm record and make an effective judgment on the alarm record; based on the different effective judgment results between different alarm records, a difference analysis is carried out between different alarm records. The clinical similarity clustering module is used to compare and analyze the changes in the clinical status of different patients, and to cluster patients of the same type based on the differences in alarm records of each alarm node. The alarm assessment and optimization module is used to extract the patient's clinical status from any alarm record and perform an effective alarm assessment on the alarm record based on comparison with similar records. Based on the valid judgment results of alarm records, the alarm thresholds for valid alarms are individually optimized; The real-time anomaly detection module is used to perform real-time alarm assessment on the patient's real-time clinical data, extract the patient's individual alarm threshold, and send alarm reminders when the individual alarm threshold is met.

7. The intensive care unit data management system according to claim 6, characterized in that: The historical alarm acquisition module includes a clinical status integration unit and an alarm record generation unit; The clinical status integration unit is used to collect historical clinical data of patients in the intensive care unit, analyze changes in the patients' clinical status and integrate them; the alarm record generation unit is used to capture alarm nodes in the process of changes in the patients' clinical status and generate corresponding alarm records.

8. The intensive care unit data management system according to claim 6, characterized in that: The alarm difference analysis module includes an alarm valid identification unit and an alarm difference analysis unit; The alarm effective identification unit is used to collect clinical intervention information for any alarm record and to make effective judgments on the alarm record; The alarm difference analysis unit is used to perform difference analysis on different alarm records based on the different valid judgment results between different alarm records.

9. The intensive care unit data management system according to claim 6, characterized in that: The alarm evaluation and optimization module includes an effective alarm evaluation unit and a personalized threshold optimization unit; The effective alarm assessment unit is used to extract the patient's clinical status from any alarm record and assess the effectiveness of the alarm record based on comparison with similar records; the personalized threshold optimization unit is used to optimize the alarm threshold of the effective alarm based on the effective judgment result of the alarm record.

Images