A method and system for predicting waiting time of outpatients in a hospital

Abstract

The present invention discloses a method and system for predicting the waiting time of outpatients in hospitals, which specifically relates to the technical field of data prediction and analysis, and is used to solve the problem of inaccurate prediction of the waiting time of outpatients in hospitals; the present invention realizes the time series analysis and labeling of the peak period characteristics of outpatients in hospitals through multi-source data collection and preprocessing, combined with external data, and accurately extracts the relevant characteristics of the peak period by using time series analysis and principal component analysis. Through the dynamic recognition system, the patient's arrival and resource load changes are monitored in real time, the peak period is quickly captured, and the waiting time benchmark value is set, the resource usage is evaluated in real time, the load is reasonably allocated, and the multi-dimensional data is combined for adaptive adjustment, the patient's waiting time is predicted, and it is flexibly adjusted according to the actual situation. Then, by analyzing the matching degree between the prediction and the actual waiting time, a feedback signal is generated, the prediction is optimized, the peak period pressure is responded to in time, and the patient experience and hospital operation efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data prediction and analysis, and more specifically, to a method and system for predicting the waiting time of outpatients in a hospital. Background Art

[0002] With the continuous growth of medical needs, the number of patients visiting hospital outpatient clinics has fluctuated. The concentrated arrival of patients during peak hours can easily lead to long waiting times, which puts pressure on patient experience and hospital operations. Long waiting times for patients may lead to a decline in medical experience, increased patient anxiety, and may even affect treatment outcomes. At the same time, irrational allocation of resources may also lead to low utilization or excessive load on medical resources such as doctors and equipment during peak hours, further exacerbating the waiting problem.

[0003] Currently, hospital outpatient departments usually rely on fixed queuing and appointment systems to manage patients' waiting times. However, these methods are lagging and inflexible when dealing with uncertain patient flow and resource load. They lack comprehensive analysis of real-time data and multidimensional factors (such as patient visits, physician workload, equipment utilization, external environmental changes, etc.), making it difficult to effectively adjust resources during peak hours, which in turn leads to inaccurate waiting time predictions. Summary of the invention

[0004] In order to overcome the above defects of the prior art, the present invention provides a method and system for predicting the waiting time of hospital outpatients, so as to solve the problem of inaccurate prediction and adjustment of the waiting time of hospital outpatients in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for predicting the waiting time of hospital outpatients comprises the following steps:

[0007] Collect and preprocess multi-source data of hospital outpatient departments, extract time series features during peak hours, combine external data, select features through time series analysis and principal component analysis, and label the features during peak hours;

[0008] Real-time analysis of patient arrival data and resource load data, using time series trend analysis to detect changes in load and equipment utilization, and dynamically identify peak hours;

[0009] During peak hours, a benchmark waiting time is set, resource usage is judged based on the comprehensive load index, and the predicted waiting time is adaptively adjusted based on multi-dimensional data, and the adjusted predicted waiting time is matched and analyzed with the actual waiting time;

[0010] The adjustment and adaptation information generated by the matching analysis process between the predicted waiting time and the actual waiting time is obtained, the adjustment and adaptation information is analyzed and different signals are generated, and the waiting time is adjusted according to the generated different signals.

[0011] In a preferred embodiment, multi-source data collection and data preprocessing of hospital outpatient departments are performed to extract time series features during peak hours, and feature selection is performed through time series analysis and principal component analysis in combination with external data. The specific process is as follows:

[0012] Data sources include historical patient visit data, doctor schedules, equipment usage, external environment data, and hospital IoT device data;

[0013] Use K-means clustering for outlier detection, divide the data into different clusters, identify points far from the cluster center as outliers, select the initial cluster center, calculate the distance between the data point and each cluster center for each data point, and assign the data point to the nearest cluster, calculate the new center of each cluster, and mark the points far from the center of each cluster as outliers;

[0014] Use interpolation or multiple imputation to complete incomplete or abnormal data;

[0015] Integrate and fuse data from different sources, that is, use outer joins and cross-matching to associate and match data;

[0016] Fast Fourier transform is used to analyze the periodic fluctuations in historical data, and principal component analysis is used to extract peak features. For external factors such as weather and traffic, natural language processing or time series correlation analysis is used to determine the influencing factors;

[0017] The features are normalized, and the time series features are segmented by time windows and data smoothed to construct composite features for representing the changes during peak hours;

[0018] Determine the interacting feature pairs and select the correlated feature pairs for combination.

[0019] In a preferred implementation, the peak period characteristics are labeled, and the specific process is as follows:

[0020] Group and organize historical data, extract and group the characteristics of patient visits, department load, and equipment utilization rate in each set time period, and select a fixed time window for segmented statistics;

[0021] According to the load fluctuation in historical data, set the load threshold during peak hours, and set the equipment utilization rate threshold based on the equipment utilization rate to adjust the label during peak hours;

[0022] According to the adjustment results of the load threshold and equipment utilization rate, each time period is marked. If the load and equipment utilization rate of the time period exceed the set load threshold and equipment utilization rate threshold, the time period is marked as a peak period, otherwise, it is marked as a non-peak period.

[0023] In a preferred embodiment, during the peak period, a waiting time benchmark value is set, resource usage is judged according to the comprehensive load index, and the predicted waiting time is adaptively adjusted in combination with multi-dimensional data, and the adjusted predicted waiting time is matched and analyzed with the actual waiting time. The specific process is as follows:

[0024] Use sliding window technology to analyze the changing trends of load and equipment utilization rate in the current period, calculate the slope of load change, determine the formation of peak period based on the slope, and issue a warning in real time when the formation of peak period is detected;

[0025] During peak hours, load detection is performed on the number of patients visiting, the real-time utilization rate of equipment, and the workload of doctors, and a comprehensive load index is calculated;

[0026] The predicted waiting time is adjusted based on the comprehensive load index, and the adjusted waiting time prediction value is determined in combination with the fluctuation of patient visits, the time period difference of equipment use, and the actual working status of doctors;

[0027] Determine the adjusted waiting time prediction value, provide real-time feedback based on the actual patient waiting situation, and determine how well the prediction matches the actual waiting time.

[0028] In a preferred embodiment, the adjustment and adaptation information generated by the matching analysis process of the predicted waiting time and the actual waiting time is obtained, and the specific process is as follows:

[0029] Obtaining adjustment and adaptation information generated in the process of predicting and verifying the error trend, the adjustment and adaptation information includes fluctuation trend information and load response information;

[0030] The fluctuation trend information includes the forecast error fluctuation trend index, and the load response information includes the load adjustment response deviation index;

[0031] The obtained forecast error fluctuation trend index and load adjustment response deviation index are combined to generate a forecast adjustment coefficient;

[0032] The forecast error fluctuation trend index and load adjustment response deviation index are proportional to the forecast adjustment coefficient.

[0033] In a preferred embodiment, the adjustment adaptation information is analyzed and different signals are generated, and the waiting time is adjusted according to the generated different signals. The specific steps are as follows:

[0034] Compare the generated forecast adjustment factor with the set waiting time adjustment threshold;

[0035] If the prediction adjustment coefficient is greater than or equal to the waiting time adjustment threshold, a time prediction adjustment signal is generated to perform an intervention adjustment on the patient's waiting time;

[0036] If the prediction adjustment coefficient is less than the waiting time adjustment threshold, a time prediction stability signal is generated and there is no need to adjust the patient's waiting time.

[0037] A hospital outpatient waiting time prediction system, used to implement the hospital outpatient waiting time prediction method, comprising:

[0038] The outpatient data collection module is used to collect and preprocess multi-source data from hospital outpatient departments, extract time series features during peak hours, and combine external data to select features through time series analysis and principal component analysis, and label the peak hour features;

[0039] Peak period identification module, which is used to analyze patient arrival data and resource load data in real time, using time series trend analysis to detect changes in load and equipment utilization, and dynamically identify peak periods;

[0040] The forecast time analysis module is used to set the waiting time benchmark value during peak hours, judge the resource usage according to the comprehensive load index, and make adaptive adjustments to the forecast waiting time based on multi-dimensional data, and perform matching analysis between the adjusted forecast waiting time and the actual waiting time;

[0041] The verification and adjustment module is used to obtain the adjustment and adaptation information generated by the matching analysis process of the predicted waiting time and the actual waiting time, analyze the adjustment and adaptation information and generate different signals, and adjust the waiting time according to the generated different signals.

[0042] Technical effects and advantages of the present invention:

[0043] The present invention realizes the time series analysis and labeling of the peak period characteristics of hospital outpatient clinics through the collection and preprocessing of multi-source data and the combination of external data, that is, the characteristic data related to the peak period is accurately extracted by using the time series analysis and principal component analysis method, and the changes of patient arrival and resource load are timely monitored through the dynamic recognition system, so that the formation of the peak period can be quickly captured. During the peak period, the waiting time benchmark value is set and the resource usage is evaluated in real time to ensure the reasonable allocation of resource load. Then, the adaptive adjustment is performed by integrating the multi-dimensional data, effectively predicting the patient's waiting time and flexibly adjusting the prediction results according to the actual situation, so as to further improve the accuracy of the prediction.

[0044] In addition, the present invention obtains adjustment and adaptation information by analyzing the match between the predicted waiting time and the actual waiting time, and generates different feedback signals. According to these signals, the prediction of patient waiting time can be dynamically optimized, thereby ensuring the continuity and accuracy of the prediction, being able to respond to the resource load pressure during the outpatient peak period in a timely manner, optimizing the patient's waiting experience, and effectively improving the hospital's operating efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 The present invention is a flow chart of a method for predicting waiting time of hospital outpatients.

[0046] Figure 2 The present invention is a schematic structural diagram of a hospital outpatient waiting time prediction system. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] Example 1: Figure 1 As shown, a method for predicting the waiting time of hospital outpatients includes the following steps:

[0049] Collect and preprocess multi-source data of hospital outpatient departments, extract time series features during peak hours, combine external data, select features through time series analysis and principal component analysis, and label the features during peak hours;

[0050] Real-time analysis of patient arrival data and resource load data, using time series trend analysis to detect changes in load and equipment utilization, and dynamically identify peak hours;

[0051] During peak hours, a benchmark waiting time is set, resource usage is judged based on the comprehensive load index, and the predicted waiting time is adaptively adjusted based on multi-dimensional data, and the adjusted predicted waiting time is matched and analyzed with the actual waiting time;

[0052] The adjustment and adaptation information generated by the matching analysis process between the predicted waiting time and the actual waiting time is obtained, the adjustment and adaptation information is analyzed and different signals are generated, and the waiting time is adjusted according to the generated different signals.

[0053] Step 1: Collect relevant information from multiple data sources of hospital outpatient departments and integrate the data. The data sources include historical patient visit data, doctor schedules, equipment usage, external environment data (such as weather and traffic conditions), hospital IoT device data, etc. The data formats and structures from different sources are different. The integration process requires data from different data sources to be standardized.

[0054] Establish data collection and storage interfaces for different data sources, import all data into a data warehouse or real-time database, use ETL (Extract, Transform, Load) tools to clean and convert data formats, and ensure data integrity and consistency;

[0055] Use data cleaning algorithms (such as regular expressions and K-means clustering for anomaly detection) and data integration tools (such as Airflow) to achieve multi-source data fusion;

[0056] In the process of data cleaning, abnormal data should be detected. K-means clustering is used for outlier detection. By dividing the data into different clusters, points far away from the cluster center are identified as outliers. K initial cluster centers are selected. For each data point x, its distance from each cluster center is calculated and assigned to the nearest cluster. The new center of each cluster is calculated, and points far away from the cluster center (points exceeding the set distance threshold) are marked as outliers.

[0057] The formula is as follows: , where k is the number of clusters, indicating that the data is divided into k clusters; represents the set of the i-th cluster, x belongs to the i-th cluster data points, is the center of the ith cluster, It represents the Euclidean distance between the data point x and the center of the i-th cluster;

[0058] Multi-source data may contain inconsistent formats (such as time, date, text fields, etc.). Regular expressions can be used to unify the format and convert all date, time and other information into a standard form.

[0059] Interpolation or multiple interpolation can be used to complete incomplete or abnormal data to ensure data quality and continuity. Interpolation assumes that data changes linearly in the previous and next time periods and is suitable for time series data with smoother trends. The specific steps of using linear interpolation are as follows: In the time series, find the nearest known value before and after the missing value. and , using the linear interpolation formula, calculate the missing values , use the interpolation results to replace the missing values ​​in the original data. The formula is as follows: ,in, and is the timestamp corresponding to the known points before and after, is the timestamp corresponding to the missing point;

[0060] For more complex missing data problems, especially in multidimensional data, multiple interpolation methods are used to handle missing data. Multiple sets of interpolation data are generated through multiple random samplings, and finally all the results are combined to obtain the completed values. The specific steps are as follows:

[0061] For each missing value, multiple interpolation data sets are generated, and interpolation calculations are performed on each data set. Assuming that the interpolation value follows the conditional distribution of the data, the results of all interpolation data sets are averaged as the final missing value completion result;

[0062] That is, randomly generate multiple sets of interpolation data (assuming there are m sets of interpolation data) ; The data for each interpolation Sampling from the conditional distribution: ,in, It represents the probability distribution of missing data under the condition of observed data. It is usually assumed to be a normal distribution or other suitable distribution. The final interpolation result is obtained by the mean value of all interpolated data. ;

[0063] After data integration and fusion, data cleaning and completion, it is necessary to integrate and fuse data from different sources, that is, to associate, match and merge the data so that multi-source data can be used in coordination with each other. The specific steps are as follows:

[0064] In data fusion, Outer Join and Cross Matching are used to integrate multiple data sources, that is, to find common fields in two data sets as connection conditions (such as timestamps, patient IDs);

[0065] Perform a full outer join on two datasets R and S. When joining, for records that have matches in R but not in S, fill the fields of S with NULL; similarly, for records that exist in S but not in R, fill the fields of R with NULL. For NULL values, you can use missing data processing strategies (such as interpolation or filling with default values); the merged dataset contains all records in R and S.

[0066] Based on the integrated data, the features related to peak hours are extracted, including the time distribution of patients’ arrival, load changes in each department, fluctuations in equipment utilization, etc. External factors such as holidays, weather changes, and traffic conditions also need to be characterized, and data of different dimensions are converted into feature vectors for prediction.

[0067] Through historical data analysis, we extract time series features related to patient visits, doctor scheduling, equipment utilization, etc., and combine external data to determine the features that affect the regularity of visits, such as the impact of weather changes on patients arriving at the hospital. The key to feature extraction is to capture the regular changes in multiple dimensions such as time points, department load, and equipment usage;

[0068] Use time series feature extraction methods (such as FFT Fourier transform) to analyze periodic fluctuations in historical data, use feature selection algorithms (such as principal component analysis PCA) to extract key peak features and eliminate redundant features. For external factors such as weather and traffic, use natural language processing (NLP) or time series correlation analysis to explore potential influencing factors;

[0069] The extracted features need to be further processed and transformed, mainly including data normalization, standardization, discretization and other operations. The numerical features in the original data are normalized so that they can be compared on the same scale. For time series features, composite features used to represent changes in peak hours are constructed through operations such as time window segmentation and data smoothing. In addition, new features can be generated by using feature interactions (such as cross features and feature combinations).

[0070] Determine the interactive feature pairs and select the correlated feature pairs for combination, such as patient arrival time and department load, equipment utilization rate and weather conditions, etc. Use correlation analysis (such as Pearson correlation coefficient) to evaluate the linear relationship between features and select the features with higher correlation for interaction;

[0071] Select features representing different behaviors and dimensions for interaction. For example, the combination of time features and load features can capture the load variation pattern in a certain time period. The formula is as follows: , where Pearson correlation coefficient is used to measure the correlation between features X and Y, the i-th data point of features X and Y, and the mean of features X and Y;

[0072] Perform combination operations on two or more features, such as multiplication, division, and logarithm, to construct a new feature by the product of two features to capture the amplification or inhibition effect between features. For example, multiply the patient's arrival time by the equipment utilization rate to generate an interactive feature that represents the equipment pressure during that period.

[0073] Label the peak time characteristics, combine the historical data of patient visits, department load, equipment utilization rate and other dimensions, and set the peak time labeling rules. The following are the specific steps:

[0074] Group and organize historical data, extract features such as the number of patient visits, department load, and equipment utilization rate in each time period, and form a comprehensive feature matrix for analyzing load fluctuations in different time periods, that is, group historical data by time units such as hours, days, and weeks. Select appropriate time windows (such as every hour, every half hour) for segmented statistics to capture changes in the number of visits and load, calculate key indicators such as the total number of patient visits, department workload, and equipment utilization rate in each time period, and normalize the statistical data of each time window to measure the relative load of the period, which is convenient for comparison with the same time period on other days;

[0075] According to the load fluctuations in historical data, set the load threshold for peak hours. Peak hours usually correspond to periods with large patient visits and department loads. Therefore, a dynamic threshold can be set to mark the time period exceeding the threshold as a peak period. For example, a dynamic threshold for peak load can be set based on the distribution of patient visits and department loads. The percentile method or the fluctuation pattern based on historical data can be used to determine the appropriate load upper limit, and the time period when the load exceeds the threshold can be marked and defined as a peak period.

[0076] In addition to the number of patients and department load, equipment utilization also plays an important role during peak hours. You can set a threshold for equipment utilization and fine-tune the labels during peak hours based on the equipment utilization to further improve the accuracy of labeling.

[0077] Statistics are collected on the equipment utilization rate in each time period to determine the equipment load. If the number of patients visiting a certain time period is high but the equipment utilization rate is low, the load label of the time period can be appropriately lowered to avoid the "pseudo peak" of idle equipment. On the contrary, if the equipment utilization rate in a certain time period is too high, the time period will be marked as a peak period.

[0078] According to the adjustment results of the load threshold and equipment utilization rate, each time period is marked. If the load and equipment utilization rate of the time period exceed the set load threshold and equipment utilization rate threshold, the time period is marked as "peak period", otherwise it is marked as "off-peak period";

[0079] It is also possible to perform label smoothing and optimization, that is, to smooth the generated labels when the peak annotations of critical points may fluctuate;

[0080] Accurate peak hours can be extracted and labeled from historical data. The labeling process combines data from multiple dimensions, such as patient visits, department load, and equipment utilization, to ensure that labels for peak hours are more accurate and dynamically adjusted.

[0081] Step 2: Dynamic peak period identification mechanism. Based on the collection and analysis of multi-dimensional data, a dynamic peak period identification mechanism is designed. This mechanism can automatically identify the formation of peak periods and issue early warnings based on real-time patient arrival data and resource load data.

[0082] To ensure the accuracy of peak period identification, the trend analysis technology of time series is combined to detect the fluctuation of patient arrival load and equipment utilization rate. By detecting rapid changes in a short period of time, the formation trend of peak period can be identified and timely judgment can be made. The sliding window technology is used to analyze the change trend of load and equipment utilization rate in the current period, calculate the slope of load change, and judge whether it has entered a rapid rise period. If the slope is large, it indicates that the peak period may be forming. The formula for the load change slope is as follows: ,in, is the time difference, usually 1 (indicating adjacent time periods);

[0083] When a peak period is detected, an early warning needs to be issued in real time to prompt managers and departments to adjust resource allocation in a timely manner (such as adding doctors, turning on more equipment, etc.). The content of the early warning may include the current load situation, the estimated peak duration, and possible resource requirements.

[0084] Step 3: Implement resource pre-scheduling and balancing mechanism. After identifying the peak period, pre-scheduling and balancing of resources are carried out. Specific measures include:

[0085] During peak hours, firstly, the real-time load detection of various resources is carried out according to the data such as the number of patients, the real-time utilization rate of equipment, and the workload of doctors. The initial waiting time benchmark value is set through the load parameters. The calculation expression of the comprehensive load index is: , where P(t), C(t), D(t), R(t), U(t), and S(t) are the number of patients arriving in time period t, the maximum capacity of the department, the number of patients in the doctor's workload, the number of patients that the doctor can handle per hour, the equipment utilization rate, and the maximum number of patients that the equipment can serve, respectively;

[0086] The predicted waiting time is adjusted based on the comprehensive load index. If L(t) exceeds the predetermined load threshold, it means that the system resources are tight and the predicted waiting time needs to be increased; if L(t) is lower than the threshold, it means that the outpatient resources are still sufficient and the waiting time is relatively short. The expression of the real-time waiting time prediction value of time period t is: ,in, is the initial waiting time benchmark value (normal waiting time under no load pressure), is the preset load threshold, which indicates the comprehensive load index when the system resources are fully loaded. is the adjustment range of the waiting time, which is adjusted according to the degree to which the load exceeds the threshold. The adjusted waiting time W(t) will be used as the basic prediction result of the real-time waiting time;

[0087] After completing the basic predicted waiting time adjustment, it is also necessary to combine the real-time data of different dimensions for adaptive waiting time adjustment. This step combines the fluctuation of patient visits, the time-based differences in equipment use, the actual working status of doctors, etc., to ensure that the waiting time prediction can be further adjusted during complex peak hours. The expression of the waiting time prediction value after adaptive adjustment is: ,in, is the predicted waiting time after adaptive adjustment, is the load adjustment factor of the i-th dimension (e.g., patient fluctuation, equipment fluctuation, physician workload fluctuation), and n is the number of adjustment dimensions;

[0088] After generating the waiting time prediction value, real-time feedback is provided based on the actual patient waiting situation to determine the matching between the prediction result and the actual waiting time. When the actual waiting time deviates from the prediction value, the parameters of the prediction algorithm are adjusted in time through the feedback mechanism to make the prediction more in line with the current situation. The adjustment and adaptation information generated by the matching analysis process between the predicted waiting time and the actual waiting time is obtained to determine the matching of the waiting prediction during the peak period. The adjustment and adaptation information includes fluctuation trend information and load response information.

[0089] The fluctuation trend information includes the prediction error fluctuation trend index and is calibrated as PEOI, and the load response information includes the load adjustment response deviation index and is calibrated as LARDI;

[0090] The forecast error fluctuation trend index is used to measure the fluctuation trend of the error between the predicted waiting time and the actual waiting time, that is, to identify whether the change direction of the forecast error in a continuous time period is consistent or trendy. By analyzing the error changes in adjacent time periods, it is evaluated whether the forecast value continues to deviate from the actual value, or whether the system has a trend deviation in the process of correcting the forecast error.

[0091] In the process of adjusting the predicted waiting time during peak hours, whether there is a cumulative effect of errors or trend fluctuations, that is, the direction and magnitude of the error change, rather than the absolute value of the error, is intended to reveal whether there is a persistent deviation in the system when adjusting the forecast, and to avoid the system adjusting the forecast value in the wrong direction for a long time. The specific functions are as follows:

[0092] In terms of directional fluctuation of errors, the forecast error fluctuation trend index measures the fluctuation direction of forecast errors in different time periods. If the error change trends in adjacent time periods are consistent (i.e., the errors continue to increase or decrease), it indicates that there may be trend deviations in certain time periods.

[0093] When the forecast error fluctuation trend index is large, it means that the forecast error of the system continues to accumulate in a certain direction, indicating that the correction of the error may be excessive or insufficient, resulting in the forecast value continuing to deviate from the actual value;

[0094] The logic for obtaining the forecast error fluctuation trend index is as follows:

[0095] Get the predicted waiting time in time periods t and t+1 and , get the length of time between the patient's arrival at the clinic and the actual visit within time period t , get the actual waiting time of patients from the time of arrival to the actual treatment in time period t+1 , calculate the waiting deviation value, the calculation expression is: , where n represents the total number of time periods, that is, the number of time periods included in the covered time window;

[0096] Get the actual number of resources available in time period t , the total number of outpatient resources in time period t , calculate the available value of resources, the calculation expression is: , based on the deviation value and resource availability value, calculate the forecast error fluctuation trend index: .

[0097] It should be noted that the predicted waiting time is obtained by calculating the current load, resources, etc. The actual waiting time can be obtained from the hospital information system, that is, the patient queue data from the hospital information system, and the actual waiting time of the patient is calculated by the timestamp difference.

[0098] The Load Adjustment Response Deviation Index (LARDI) indicates whether the adjustment range of the predicted waiting time is reasonable when responding to load changes during peak hours. Specifically, it refers to whether the adjustment of the predicted waiting time matches the actual load changes. By analyzing the adjustment range of the predicted value and the load change, it is judged whether an appropriate response is made to the load change to ensure that the predicted waiting time is not over- or under-adjusted.

[0099] The load adjustment response deviation index mainly measures the rationality of the system's adjustment of the forecast waiting time when the load fluctuates, determines whether the adjustment of the forecast value is consistent with the load change, and the system's response effect to load fluctuations. It reflects whether the system can quickly and accurately adjust the forecast value when the load fluctuates violently to avoid excessive or insufficient adjustment;

[0100] When the load fluctuates, the outpatient waiting time forecast adjustment may be excessive (i.e., the forecast time is adjusted too much) or insufficient (i.e., the adjustment is not enough). The load adjustment response deviation index can measure whether the forecast value is reasonable to prevent over-response or under-response. The load adjustment response deviation index can predict the accuracy of the adjustment when monitoring the load changes, ensuring that the system can appropriately adjust the forecast value according to the load changes, rather than simply adjusting it based on sensitivity.

[0101] The logic for obtaining the load adjustment response deviation index is as follows:

[0102] Get the resource usage of the recorded time periods st and st−1 and , calculate resource utilization: , set the hysteresis time window , collection time period and The load data are recorded as and , calculate the delayed load value: , calculate the resource synergy effect value, the calculation expression is: , where is the available quantity of different resources i in time period st, is the actual required quantity of different resources i in time period st, m is the number of resource types, and the load adjustment response deviation index is calculated. The calculation expression is: , T represents the number of windows.

[0103] It should be noted that the usage data of various resources (such as doctors, equipment, and treatment rooms) are obtained from the hospital resource management system, including the working hours and occupancy rate of doctors, the utilization rate of key medical equipment (such as CT and MRI), and the idle and occupied status of treatment rooms to determine the use of outpatient resources; the resource synergy effect value reflects the synergy efficiency between different resources (such as doctors and equipment, equipment and treatment rooms). It is used to evaluate the impact of the interdependence between resources on the system prediction and adjustment. For example, when doctors are available but equipment is insufficient, the diagnosis and treatment efficiency will decrease, resulting in the failure of prediction and adjustment. By analyzing the utilization rate and work schedule between different resources, their synergy efficiency is evaluated. The higher the resource synergy, the better the synergy effect between resources, and the more accurate the response to prediction and adjustment.

[0104] The fluctuation trend information and load response information are combined to generate the forecast adjustment coefficient, that is, the obtained forecast error fluctuation trend index and load adjustment response deviation index are combined to generate the forecast adjustment coefficient. The expression is: , where , is the preset proportional coefficient of the forecast error fluctuation trend index and the load adjustment response deviation index, and , Both are greater than 0.

[0105] The specific method of jointly generating the prediction adjustment coefficient may involve multiple algorithms and models, which depends on the actual situation and application requirements. The present embodiment may use a weighted summation method to combine the prediction error fluctuation trend index and the load adjustment response deviation index to generate a comprehensive prediction adjustment coefficient. This prediction adjustment coefficient can be used as an input parameter for the prediction analysis of the waiting time of outpatient patients to determine the accuracy of the waiting time prediction.

[0106] It should be noted that the size of the preset proportional coefficient is a specific value obtained by quantifying each parameter. In order to facilitate subsequent comparison, the size of the coefficient depends on the amount of sample data and the preliminary setting of the corresponding preset proportional coefficient for each group of sample data by technical personnel in this field. It is not unique, as long as it does not affect the proportional relationship between the parameter and the quantized value. For example, the prediction error fluctuation trend index is proportional to the prediction adjustment coefficient. The prediction error fluctuation trend index and the load adjustment response deviation index are normalized to have the same dimension and range. This can be achieved by subtracting the mean from the original data and dividing it by the standard deviation, or mapping the data to the range of [0, 1].

[0107] The larger the forecast error fluctuation trend index and the larger the load adjustment response deviation index, the larger the forecast adjustment coefficient generated jointly, indicating that the cumulative effect of the forecast error in adjacent time periods is obvious, indicating that the waiting time forecast deviation has not been corrected in time. A larger forecast error fluctuation trend index indicates that continuous forecast errors have occurred in multiple time periods, requiring more complex adjustments. Higher agility and accuracy are required when adjusting forecasts, especially during peak hours or when resources are tight. When the forecast adjustment coefficient is large, it means that the waiting time forecast strategy must be adjusted faster to avoid over-response or under-response problems.

[0108] The smaller the forecast error fluctuation trend index and the load adjustment response deviation index are, the smaller the forecast adjustment coefficient generated by the combined method is. When responding to actual load changes and resource fluctuations, the required forecast adjustment amplitude is smaller, which can maintain a high forecast accuracy and reduce the additional adjustment burden. The forecast is more stable and does not require frequent or large adjustments, resulting in less forecast adjustment pressure.

[0109] Comparing the generated prediction adjustment coefficient with a preset waiting time adjustment threshold to generate a time prediction adjustment signal and a time prediction stabilization signal;

[0110] After obtaining the forecast adjustment coefficient, the forecast adjustment coefficient is compared with the waiting time adjustment threshold;

[0111] If the prediction adjustment coefficient is greater than or equal to the waiting time adjustment threshold, a time prediction adjustment signal is generated, which indicates that the outpatient patient visit volume, doctor load, equipment utilization rate and other resource conditions have changed dramatically, and these changes have not been effectively responded to. The time prediction adjustment signal generated at this time reflects the instability of the current load and resource conditions, reminding the system that the predicted waiting time must be dynamically adjusted to keep up with the changes, and timely patient waiting time intervention adjustments must be made. Measures need to be taken to rebalance resources to ensure smoother outpatient patient mobility;

[0112] If the prediction adjustment coefficient is less than the waiting time adjustment threshold, a time prediction stability signal is generated, which indicates that the system performs well in predicting the waiting time of outpatient patients, the prediction error of the system is less than the acceptable threshold, and the prediction result is consistent with the actual situation. At this time, the system does not need to make significant adjustments to the patient waiting time, the outpatient operation is stable, and the match between resources and patient flow is good.

[0113] After generating the time forecast adjustment signal, the system needs to respond quickly to adjust the waiting time forecast for outpatients, improve forecast accuracy, and cope with high load or insufficient resources. The following are specific quick response measures;

[0114] If the forecast adjustment signal indicates a surge in patient traffic, more doctors, nurses or auxiliary staff should be dispatched to busy departments immediately to share the patient load. Temporary scheduling adjustments can be made to mobilize idle medical personnel within the hospital; key equipment (such as CT, MRI, and ultrasound equipment) can be prioritized to extend the equipment's operating time or speed up the equipment's turnover, ensuring that patients can complete diagnosis and treatment procedures faster and reduce waiting time;

[0115] Re-adjust the queue order according to the severity of the patient's condition, give priority to patients who need to see a doctor urgently, so as to avoid high-risk patients waiting too long; adjust the opening hours of the appointment window in real time, reduce the number of appointments during peak hours, and disperse the patient's arrival time. Consider notifying some scheduled patients to postpone their appointment time to relieve the current load pressure;

[0116] It should be noted that the thresholds involved in the embodiments can be determined according to specific scenarios and requirements.

[0117] The present invention realizes the time series analysis and labeling of the peak period characteristics of hospital outpatient clinics through the collection and preprocessing of multi-source data and the combination of external data, that is, the characteristic data related to the peak period is accurately extracted by using the time series analysis and principal component analysis method, and the changes of patient arrival and resource load are timely monitored through the dynamic recognition system, so that the formation of the peak period can be quickly captured. During the peak period, the waiting time benchmark value is set and the resource usage is evaluated in real time to ensure the reasonable allocation of resource load. Then, the adaptive adjustment is performed by integrating the multi-dimensional data, effectively predicting the patient's waiting time and flexibly adjusting the prediction results according to the actual situation, so as to further improve the accuracy of the prediction.

[0118] In addition, the present invention obtains adjustment and adaptation information by analyzing the match between the predicted waiting time and the actual waiting time, and generates different feedback signals. According to these signals, the prediction of patient waiting time can be dynamically optimized, thereby ensuring the continuity and accuracy of the prediction, being able to respond to the resource load pressure during the outpatient peak period in a timely manner, optimizing the patient's waiting experience, and effectively improving the hospital's operating efficiency.

[0119] Embodiment 2: A hospital outpatient waiting time prediction system, such as Figure 2 As shown, specifically including:

[0120] The outpatient data collection module is used to collect and preprocess multi-source data from hospital outpatient departments, extract time series features during peak hours, and combine external data to select features through time series analysis and principal component analysis, and label the peak hour features;

[0121] Peak period identification module, which is used to analyze patient arrival data and resource load data in real time, using time series trend analysis to detect changes in load and equipment utilization, and dynamically identify peak periods;

[0122] The forecast time analysis module is used to set the waiting time benchmark value during peak hours, judge the resource usage according to the comprehensive load index, and make adaptive adjustments to the forecast waiting time based on multi-dimensional data, and perform matching analysis between the adjusted forecast waiting time and the actual waiting time;

[0123] The verification and adjustment module is used to obtain the adjustment and adaptation information generated by the matching analysis process of the predicted waiting time and the actual waiting time, analyze the adjustment and adaptation information and generate different signals, and adjust the waiting time according to the generated different signals.

[0124] The above formulas are all dimensionless and calculated numerically. Specific dimension removal can be achieved by various means such as standardization, which will not be elaborated here. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0125] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, an ATA hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium may be a solid-state ATA hard disk.

[0126] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0127] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0128] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0129] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0130] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0131] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for predicting the waiting time of outpatients in a hospital, characterized in that: The steps include: Collect and preprocess multi-source data of hospital outpatient departments, extract time series features during peak hours, combine external data, select features through time series analysis and principal component analysis, and label the features during peak hours; Real-time analysis of patient arrival data and resource load data, using time series trend analysis to detect changes in load and equipment utilization, and dynamically identify peak hours; During peak hours, a benchmark waiting time is set, resource usage is judged based on the comprehensive load index, and the predicted waiting time is adaptively adjusted based on multi-dimensional data, and the adjusted predicted waiting time is matched and analyzed with the actual waiting time; Acquire adjustment and adaptation information generated by the matching analysis process between the predicted waiting time and the actual waiting time, analyze the adjustment and adaptation information and generate different signals, and adjust the waiting time according to the generated different signals; During peak hours, a benchmark waiting time is set, resource usage is judged based on the comprehensive load index, and the predicted waiting time is adaptively adjusted based on multi-dimensional data. The adjusted predicted waiting time is matched with the actual waiting time. The specific process is as follows: Use sliding window technology to analyze the changing trends of load and equipment utilization rate in the current period, calculate the slope of load change, determine the formation of peak period based on the slope, and issue a warning in real time when the formation of peak period is detected; During peak hours, load detection is performed on the number of patients visiting, the real-time utilization rate of equipment, and the workload of doctors, and the comprehensive load index is calculated. The calculation expression is: , where P(t), C(t), D(t), R(t), U(t), and S(t) are the number of patients arriving in time period t, the maximum capacity of the department, the number of patients in the doctor's workload, the number of patients that the doctor can handle per hour, the equipment utilization rate, and the maximum number of patients that the equipment can serve, respectively; The predicted waiting time is adjusted based on the comprehensive load index, and the adjusted waiting time prediction value is determined in combination with the fluctuation of patient visits, the time period difference of equipment use, and the actual working status of doctors; Determine the adjusted waiting time prediction value, provide real-time feedback based on the actual patient waiting situation, and determine how well the prediction matches the actual waiting time.

2. A method for predicting waiting time of outpatients in a hospital according to claim 1, characterized in that: Multi-source data collection and data preprocessing of hospital outpatient departments were carried out to extract the time series characteristics of peak hours. In combination with external data, feature selection was performed through time series analysis and principal component analysis. The specific process is as follows: Data sources include historical patient visit data, doctor schedules, equipment usage, external environment data, and hospital IoT device data; Use K-means clustering for outlier detection, divide the data into different clusters, identify points far from the cluster center as outliers, select the initial cluster center, calculate the distance between the data point and each cluster center for each data point, and assign the data point to the nearest cluster, calculate the new center of each cluster, and mark the points far from the center of each cluster as outliers; Use interpolation or multiple imputation to complete incomplete or abnormal data; Integrate and fuse data from different sources, that is, use outer joins and cross-matching to associate and match data; Fast Fourier transform is used to analyze the periodic fluctuations in historical data, and principal component analysis is used to extract peak features. For external factors such as weather and traffic, natural language processing or time series correlation analysis is used to determine the influencing factors; The features are normalized, and the time series features are segmented by time windows and data smoothed to construct composite features for representing the changes during peak hours; Determine the interacting feature pairs and select the correlated feature pairs for combination.

3. A method for predicting waiting time of outpatients in a hospital according to claim 2, characterized in that: And label the peak period characteristics. The specific process is as follows: Group and organize historical data, extract and group the characteristics of patient visits, department load, and equipment utilization rate in each set time period, and select a fixed time window for segmented statistics; According to the load fluctuation in historical data, set the load threshold during peak hours, and set the equipment utilization rate threshold based on the equipment utilization rate to adjust the label during peak hours; According to the adjustment results of the load threshold and equipment utilization rate, each time period is marked. If the load and equipment utilization rate of the time period exceed the set load threshold and equipment utilization rate threshold, the time period is marked as a peak period, otherwise, it is marked as a non-peak period.

4. A method for predicting waiting time of outpatients in a hospital according to claim 1, characterized in that: Obtain the adjustment and adaptation information generated by the matching analysis process of the predicted waiting time and the actual waiting time. The specific process is as follows: Obtaining adjustment and adaptation information generated in the process of predicting and verifying the error trend, the adjustment and adaptation information includes fluctuation trend information and load response information; The fluctuation trend information includes the forecast error fluctuation trend index, and the load response information includes the load adjustment response deviation index; The obtained forecast error fluctuation trend index and load adjustment response deviation index are combined to generate a forecast adjustment coefficient; The forecast error fluctuation trend index and load adjustment response deviation index are proportional to the forecast adjustment coefficient.

5. A method for predicting waiting time of outpatients in a hospital according to claim 4, characterized in that: The adjustment adaptation information is analyzed and different signals are generated, and the waiting time is adjusted according to the generated different signals. The specific steps are as follows: Compare the generated forecast adjustment factor with the set waiting time adjustment threshold; If the prediction adjustment coefficient is greater than or equal to the waiting time adjustment threshold, a time prediction adjustment signal is generated to perform an intervention adjustment on the patient's waiting time; If the prediction adjustment coefficient is less than the waiting time adjustment threshold, a time prediction stability signal is generated and there is no need to adjust the patient's waiting time.

6. A hospital outpatient waiting time prediction system, used to implement a hospital outpatient waiting time prediction method according to any one of claims 1 to 5, characterized in that: include: The outpatient data collection module is used to collect and preprocess multi-source data from hospital outpatient departments, extract time series features during peak hours, and combine external data to select features through time series analysis and principal component analysis, and label the peak hour features; Peak period identification module, which is used to analyze patient arrival data and resource load data in real time, using time series trend analysis to detect changes in load and equipment utilization, and dynamically identify peak periods; The forecast time analysis module is used to set the waiting time benchmark value during peak hours, judge the resource usage according to the comprehensive load index, and make adaptive adjustments to the forecast waiting time based on multi-dimensional data, and perform matching analysis between the adjusted forecast waiting time and the actual waiting time; A verification and adjustment module is used to obtain adjustment and adaptation information generated by the matching analysis process between the predicted waiting time and the actual waiting time, analyze the adjustment and adaptation information and generate different signals, and adjust the waiting time according to the generated different signals; During peak hours, a benchmark waiting time is set, resource usage is judged based on the comprehensive load index, and the predicted waiting time is adaptively adjusted based on multi-dimensional data. The adjusted predicted waiting time is matched with the actual waiting time. The specific process is as follows: Use sliding window technology to analyze the changing trends of load and equipment utilization rate in the current period, calculate the slope of load change, determine the formation of peak period based on the slope, and issue a warning in real time when the formation of peak period is detected; During peak hours, load detection is performed on the number of patients visiting, the real-time utilization rate of equipment, and the workload of doctors, and the comprehensive load index is calculated. The calculation expression is: , where P(t), C(t), D(t), R(t), U(t), and S(t) are the number of patients arriving in time period t, the maximum capacity of the department, the number of patients in the doctor's workload, the number of patients that the doctor can handle per hour, the equipment utilization rate, and the maximum number of patients that the equipment can serve, respectively; The predicted waiting time is adjusted based on the comprehensive load index, and the adjusted waiting time prediction value is determined in combination with the fluctuation of patient visits, the time period difference of equipment use, and the actual working status of doctors; Determine the adjusted waiting time prediction value, provide real-time feedback based on the actual patient waiting situation, and determine how well the prediction matches the actual waiting time.

Images