A method and system for precise follow-up of chronic diseases based on big data

Abstract

This invention provides a method and system for precise follow-up of chronic diseases based on big data, relating to the field of chronic disease follow-up technology. The method includes: acquiring discharge diagnosis and treatment data and discharge vital sign data; acquiring home vital sign data; determining measured vital sign recovery data; inputting the discharge diagnosis and treatment data and the discharge vital sign data into a trained vital sign recovery prediction model to obtain predicted vital sign recovery data for multiple patients at multiple home times; determining the highest recovery coefficient based on the measured vital sign recovery data and the predicted vital sign recovery data; acquiring the ages of multiple patients, as well as the number of acute exacerbations and hospitalization frequency within one year; determining a risk coefficient based on the age, the number of acute exacerbations, and the hospitalization frequency; and determining the follow-up order and follow-up frequency. According to this invention, personalized follow-up plans can be formulated based on the highest recovery coefficient and the risk coefficient, improving the accuracy and effectiveness of follow-up and reducing the waste of medical resources.

Description

Technical Field

[0001] This invention relates to the field of chronic disease follow-up technology, and in particular to a method and system for accurate chronic disease follow-up based on big data. Background Technology

[0002] Current technologies, while enabling real-time monitoring of chronic diseases based on patients' conditions and lifestyles, do not consider the impact of patients' post-discharge recovery and chronic disease risk on follow-up. In other words, they cannot develop personalized follow-up plans based on the highest recovery coefficient and risk coefficient.

[0003] The information disclosed in the background section of this application is intended only to enhance the understanding of the general background of this application and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0004] This invention provides a method and system for precise follow-up of chronic diseases based on big data, which can solve the technical problem that related technologies cannot formulate personalized follow-up plans based on the highest recovery coefficient and risk coefficient.

[0005] According to a first aspect of the present invention, a method for precise follow-up of chronic diseases based on big data is provided, comprising: acquiring discharge diagnosis and treatment data and discharge vital sign data of multiple patients, wherein the discharge diagnosis and treatment data includes total discharge treatment cost, length of hospital stay and chronic disease type, and the discharge vital sign data includes discharge weight and multiple discharge physiological indicators; acquiring home vital sign data of multiple patients at multiple home times, wherein the home vital sign data includes home weight and multiple home physiological indicators; determining measured vital sign rehabilitation data based on the home vital sign data, standard weight and multiple standard physiological indicators; inputting the discharge diagnosis and treatment data and the discharge vital sign data into a trained vital sign rehabilitation prediction model to obtain predicted vital sign rehabilitation data of multiple patients at multiple home times; determining the highest rehabilitation coefficient based on the measured vital sign rehabilitation data and the predicted vital sign rehabilitation data; acquiring the age of multiple patients, as well as the number of acute exacerbations and hospitalization frequency within one year; determining a risk coefficient based on the age, the number of acute exacerbations and the hospitalization frequency; and determining the follow-up order and follow-up frequency based on the highest rehabilitation coefficient and the risk coefficient.

[0006] Further, based on the home-based vital sign data, standard weight, and multiple standard physiological indicator data, the measured vital sign rehabilitation data are determined, including: obtaining home-based vital sign vectors for multiple patients at multiple home-based times based on home-based vital sign data for multiple patients at multiple home-based times; obtaining standard vital sign vectors based on the standard weight and the standard physiological indicator data; and determining the similarity between the home-based vital sign vectors and the standard vital sign vectors as the measured vital sign rehabilitation data for the j-th patient at the k-th home-based time, where k and j are both positive integers.

[0007] Furthermore, the training steps of the vital sign rehabilitation prediction model include: acquiring historical discharge treatment data and historical discharge vital sign data of multiple historical patients, wherein the historical discharge treatment data includes historical total treatment cost, historical length of hospital stay, and historical chronic disease type, and the historical discharge vital sign data includes historical discharge weight and multiple historical discharge physiological indicators; acquiring historical home vital sign data of historical patients at multiple historical home times, wherein the historical home vital sign data includes historical home weight and multiple historical home physiological indicators; and based on the historical home vital sign data, standard weight, and standard physiological indicators... The system identifies historical measured vital sign rehabilitation data; it processes the historical discharge treatment data and historical discharge vital sign data using a vital sign rehabilitation prediction model to obtain historical predicted vital sign rehabilitation data for multiple patients at multiple historical home care times; it obtains the historical age of the historical patients; it determines the loss function of the vital sign rehabilitation prediction model based on the historical age, the historical discharge treatment data, the historical measured vital sign rehabilitation data, and the historical predicted vital sign rehabilitation data; and it trains the vital sign rehabilitation prediction model based on the loss function to obtain the trained vital sign rehabilitation prediction model.

[0008] Further, based on the historical age, the historical discharge and treatment data, the historical measured vital sign rehabilitation data, and the historical predicted vital sign rehabilitation data, the loss function of the vital sign rehabilitation prediction model is determined, including: according to the formula Determine the loss function LOSS of the vital sign rehabilitation prediction model, where, This represents the historical measured vital signs and rehabilitation data of the i-th historical patient at the h-th historical home-staying time. This provides the historical predicted vital sign recovery data for the i-th historical patient at the h-th historical home-based time point. Let be the number of days of hospitalization for the i-th historical patient. Standard length of stay, Let i be the total historical discharge treatment cost of the i-th historical patient. The standard total treatment price, For the i-th historical patient, the type of historical chronic disease. For the j-th patient, the type of chronic disease is... Let be the historical age of the i-th historical patient. Let S be the age of the j-th patient, J be the number of patients, S be the number of physiological indicators, and H be the number of historical home-based time periods. Let N be the number of historical patients in the e-th training batch, and N be the number of training batches, where j ≤ J, h ≤ H, and i ≤ 1. , e≤N, and j, a, h, x, i, J, H, Both N and N are positive integers.

[0009] Further, based on the measured vital sign rehabilitation data and the predicted vital sign rehabilitation data, the highest rehabilitation coefficient is determined, including: determining the rehabilitation coefficients of multiple patients at multiple home-based times based on the measured vital sign rehabilitation data and the predicted vital sign rehabilitation data; and determining the highest rehabilitation coefficient for each patient based on the maximum value of the rehabilitation coefficients of each patient at multiple home-based times.

[0010] Furthermore, based on the measured vital sign recovery data and the predicted vital sign recovery data, the recovery coefficients of multiple patients at multiple home times are determined, including: according to the formula Determine the recovery coefficient of the j-th patient at the k-th home time. ,in, This represents the measured vital signs and rehabilitation data of the j-th patient at the k-th home time. This is the predicted vital sign recovery data for the j-th patient at the k-th home time, where K is the number of home time, k≤K, and both k and K are positive integers, and if is a conditional function.

[0011] Further, the risk coefficient is determined based on the age, the number of acute attacks, and the hospitalization frequency, including: obtaining the frequency of family member involvement in care through nursing records; if the frequency of family member involvement in care is 3 times or more per week, the patient comfort outcome is determined to be 1; if the frequency of family member involvement in care is less than 3 times per week, the patient comfort outcome is determined to be 0; and the risk coefficient is determined based on the patient comfort outcome, the age, the number of acute attacks, and the hospitalization frequency.

[0012] Further, a risk factor is determined based on the patient's comfort outcome, age, number of acute exacerbations, and frequency of hospitalizations, including: according to the formula Determine the risk coefficient for the j-th patient. ,in, Let j be the age of the j-th patient. Let j be the number of acute attacks for the j-th patient. Let j be the hospitalization frequency of the j-th patient. For patient comfort outcomes, , and The preset weights are used, and max is the function to find the maximum value.

[0013] Further, determining the follow-up order and frequency based on the highest recovery coefficient and the risk coefficient includes: arranging the highest recovery coefficients of multiple patients in ascending order to determine the follow-up sequence; determining the follow-up order based on the follow-up sequence; if the risk coefficient is greater than or equal to a preset risk coefficient, then determining the follow-up frequency as weekly; if the risk coefficient is less than the preset risk coefficient, then determining the follow-up frequency as monthly.

[0014] According to a second aspect of the present invention, a big data-based system for precise follow-up of chronic diseases is provided, comprising: a discharge diagnosis and treatment data and discharge vital sign data module, used to acquire discharge diagnosis and treatment data and discharge vital sign data of multiple patients, wherein the discharge diagnosis and treatment data includes total discharge treatment cost, length of hospital stay and chronic disease type, and the discharge vital sign data includes discharge weight and multiple discharge physiological indicators; a home vital sign data module, used to acquire home vital sign data of multiple patients at multiple home times, wherein the home vital sign data includes home weight and multiple home physiological indicators; and a measured vital sign rehabilitation data module, used to determine measured vital sign rehabilitation data based on the home vital sign data, standard weight and multiple standard physiological indicators. The system includes: a predicted vital sign rehabilitation data module, used to input the discharge diagnosis and treatment data and the discharge vital sign data into a trained vital sign rehabilitation prediction model to obtain predicted vital sign rehabilitation data for multiple patients at multiple home times; a maximum rehabilitation coefficient module, used to determine the maximum rehabilitation coefficient based on the measured vital sign rehabilitation data and the predicted vital sign rehabilitation data; an age, number of acute attacks, and hospitalization frequency module, used to obtain the age of multiple patients, as well as the number of acute attacks and hospitalization frequency within one year; a risk coefficient module, used to determine the risk coefficient based on the age, the number of acute attacks, and the hospitalization frequency; and a follow-up order and follow-up frequency module, used to determine the follow-up order and follow-up frequency based on the maximum rehabilitation coefficient and the risk coefficient.

[0015] Technical effects: According to the present invention, discharge diagnosis and treatment data, discharge vital sign data, and home vital sign data help to more accurately assess the patient's health status. Using a trained vital sign rehabilitation prediction model, the predicted vital sign data of the patient at multiple home moments can be predicted, which helps to quantitatively assess the patient's rehabilitation progress and effect. The highest rehabilitation coefficient can more accurately identify the priority of follow-up, and the risk coefficient helps to identify high-risk patients. That is, personalized follow-up plans can be formulated based on the highest rehabilitation coefficient and the risk coefficient, improving the accuracy and effectiveness of follow-up and reducing the waste of medical resources. When determining the loss function of the vital signs rehabilitation prediction model, the impact of historical hospital stays and historical total treatment costs on the severity of the patient's condition can be used to determine the influence of these data on the error of historical vital signs rehabilitation data. Based on this influence and the relative difference between historical measured vital signs rehabilitation data and historical predicted vital signs rehabilitation data, weights are set based on the patient's chronic disease type and age. Weights are also set based on the characteristics that the more similar the patient's sample has to historical chronic disease type and age, the greater the reference value. Furthermore, weights are set based on the characteristics that the shorter the time interval from the start of home isolation, the higher the accuracy, and the shorter the time interval from the first batch, the lower the accuracy. In this way, the errors output by the vital signs rehabilitation prediction model for multiple historical patients at multiple historical home isolation times in each training batch are weighted and summed to obtain the loss function. This improves the design accuracy and objectivity of the loss function, thereby improving training efficiency and the accuracy of the vital signs rehabilitation prediction model during the training process. When determining the recovery coefficient for multiple patients at multiple home-based times, weights can be assigned based on the characteristic that shorter time intervals from the start of home-based care result in higher accuracy. This allows for a reasonable weighting of the recovery degree errors at different home-based times, yielding a recovery coefficient that reflects the deviation between the patient's actual and expected recovery status, thus improving the accuracy and reliability of the recovery coefficient. When determining the risk coefficient, multiple factors can be considered, including the patient's age, number of acute exacerbations, frequency of hospitalizations, and comfort outcomes. Weights can be assigned based on the characteristic that older patients have a higher risk of relapse. Determining the risk coefficient allows for understanding the risk level of each patient, facilitating the development of more precise follow-up plans and providing differentiated follow-up services for patients with different risk levels.

[0016] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Other features and aspects of the invention will become clearer from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained based on these drawings without creative effort.

[0018] Figure 1 An exemplary flowchart of a big data-based method for precise follow-up of chronic diseases according to an embodiment of the present invention is shown.

[0019] Figure 2 An exemplary flowchart illustrating the calculation of measured vital sign rehabilitation data according to an embodiment of the present invention is shown;

[0020] Figure 3 An exemplary flowchart for calculating the highest recovery coefficient according to an embodiment of the present invention is shown;

[0021] Figure 4 A flowchart for calculating the risk coefficient according to an embodiment of the present invention is shown as an example;

[0022] Figure 5 A flowchart illustrating the arrangement of follow-up sequence and frequency according to an embodiment of the present invention is shown by way of example;

[0023] Figure 6 A block diagram of a big data-based precision follow-up system for chronic diseases is shown as an example according to an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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.

[0025] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0026] Figure 1An exemplary flowchart of a big data-based method for precise follow-up of chronic diseases according to an embodiment of the present invention is shown. The method includes: Step S1, acquiring discharge medical data and discharge vital sign data of multiple patients, wherein the discharge medical data includes total discharge treatment cost, length of hospital stay, and type of chronic disease, and the discharge vital sign data includes discharge weight and multiple discharge physiological indicators; Step S2, acquiring home vital sign data of multiple patients at multiple home times, wherein the home vital sign data includes home weight and multiple home physiological indicators; Step S3, based on the home vital sign data, standard weight, and multiple standard physiological indicators... According to the measured vital sign rehabilitation data, step S4, the discharge treatment data and the discharge vital sign data are input into the trained vital sign rehabilitation prediction model to obtain predicted vital sign rehabilitation data for multiple patients at multiple home times; step S5, the highest rehabilitation coefficient is determined based on the measured vital sign rehabilitation data and the predicted vital sign rehabilitation data; step S6, the ages of multiple patients, as well as the number of acute attacks and the frequency of hospitalization within one year are obtained; step S7, the risk coefficient is determined based on the age, the number of acute attacks, and the frequency of hospitalization; step S8, the follow-up order and follow-up frequency are determined based on the highest rehabilitation coefficient and the risk coefficient.

[0027] The big data-based method for precise follow-up of chronic diseases according to embodiments of the present invention, through discharge diagnosis and treatment data, discharge vital sign data, and home vital sign data, helps to more accurately assess the patient's health status. Utilizing a trained vital sign rehabilitation prediction model, it can predict the patient's predicted vital sign data at multiple home moments, which helps to quantitatively assess the patient's rehabilitation progress and effects. The highest rehabilitation coefficient can more accurately identify the priority of follow-up, and the risk coefficient helps to identify high-risk patients. That is, personalized follow-up plans can be formulated based on the highest rehabilitation coefficient and the risk coefficient, improving the accuracy and effectiveness of follow-up and reducing the waste of medical resources.

[0028] According to one embodiment of the present invention, in step S1, the patient's most recent discharge medical data is obtained based on the patient's most recent discharge record. This includes recording the total treatment cost for each patient upon discharge, i.e., the sum of all treatment costs incurred from admission to discharge, as well as the number of days of hospitalization. Furthermore, the patient's chronic disease type needs to be identified, such as hypertension, diabetes, coronary heart disease, chronic hepatitis, etc. Upon discharge, the patient is weighed to obtain their discharge weight, and multiple physiological indicators are checked, such as blood pressure, blood sugar, and heart rate.

[0029] According to one embodiment of the present invention, in step S2, the interval between adjacent home times can be set to 12 hours, 24 hours, etc., and the present invention does not limit this. After the patient is discharged from the hospital, it is necessary to collect home vital sign data at multiple home times during their home life. For example, using a scale, the patient's home weight data can be accurately obtained, and multiple home physiological indicator data can be collected, such as blood pressure measurement values, blood glucose test results, heart rate values, etc., in the home environment, which are consistent with the physiological indicator content of the discharge physiological indicator data.

[0030] According to one embodiment of the present invention, in step S3, measured physical sign rehabilitation data are determined based on the home vital sign data, standard weight and multiple standard physiological indicator data.

[0031] Figure 2 A flowchart illustrating the calculation of measured vital sign rehabilitation data according to an embodiment of the present invention is shown as an example.

[0032] According to an embodiment of the present invention, step S3 includes: step S31, obtaining home vital sign vectors for multiple patients at multiple home times based on home vital sign data of multiple patients at multiple home times; step S33, obtaining standard vital sign vectors based on the standard weight and the standard physiological index data; step S33, determining the similarity between the home vital sign vectors and the standard vital sign vectors as the measured vital sign rehabilitation data of the j-th patient at the k-th home time, where k and j are both positive integers.

[0033] According to one embodiment of the present invention, home vital sign data, namely home weight and multiple home physiological indicator data, can be combined into a home vital sign vector, and standard weight and multiple standard physiological indicator data can be combined into a standard vital sign vector, based on the calculation formula of measured vital sign rehabilitation data. ,in, This represents the measured vital signs and rehabilitation data of the j-th patient at the k-th home time. Let J be the home-based vital signs vector of the j-th patient at the k-th home-based time. For standard trait vectors, for The transpose of the vector, where k and j are both positive integers. The larger the measured vital signs rehabilitation data, the closer the patient's home vital signs data are to the standard weight and standard physiological indicators at that time, that is, the better the patient's rehabilitation effect.

[0034] According to an embodiment of the present invention, in step S4, discharge diagnosis and treatment data and discharge vital sign data are input into a trained vital sign rehabilitation prediction model. The vital sign rehabilitation prediction model can be a convolutional neural network model, which is trained based on a large amount of sample data through machine learning or deep learning technology to obtain predicted vital sign rehabilitation data of multiple patients at multiple home times.

[0035] According to one embodiment of the present invention, the training steps of the vital sign rehabilitation prediction model include: acquiring historical discharge medical data and historical discharge vital sign data of multiple historical patients, wherein the historical discharge medical data includes historical total treatment cost at discharge, historical length of hospital stay, and historical chronic disease type, and the historical discharge vital sign data includes historical discharge weight and multiple historical discharge physiological indicators; acquiring historical home vital sign data of historical patients at multiple historical home times, wherein the historical home vital sign data includes historical home weight and multiple historical home physiological indicators; and based on the historical home vital sign data, standard weight, and standard living... The system analyzes and determines historical measured vital sign rehabilitation data using indicator data. It then processes the historical discharge treatment data and the historical discharge vital sign data using a vital sign rehabilitation prediction model to obtain historical predicted vital sign rehabilitation data for multiple patients at multiple historical home-based times. The system also obtains the historical age of each patient. Based on the historical age, the historical discharge treatment data, the historical measured vital sign rehabilitation data, and the historical predicted vital sign rehabilitation data, the system determines the loss function of the vital sign rehabilitation prediction model. Finally, it trains the vital sign rehabilitation prediction model using the loss function to obtain the trained vital sign rehabilitation prediction model.

[0036] According to one embodiment of the present invention, by utilizing medical historical big data, historical discharge treatment data and historical discharge vital sign data of multiple patients with chronic diseases are obtained. These patients are then divided into different training batches, with each training batch containing the same number of patients. This allows for the training of a vital sign rehabilitation prediction model across different training batches. For example... M represents the number of historical patients. , , ..., These represent the number of historical patients in training batches 1, 2, ..., N. The method for obtaining historical measured vital sign rehabilitation data is similar to that for obtaining measured vital sign rehabilitation data, and will not be repeated here. Higher treatment costs and longer hospital stays indicate more severe patient conditions and slower post-discharge vital sign rehabilitation, meaning smaller predicted vital sign rehabilitation data. The vital sign rehabilitation prediction model can predict historical predicted vital sign rehabilitation data for multiple historical patients at multiple historical home-based times based on the relationship between treatment costs, hospital stays, and disease severity, and based on historical discharge treatment data and historical discharge vital sign data. The loss function is determined based on the difference between historical measured vital sign rehabilitation data and historical predicted vital sign rehabilitation data. By adjusting the loss function through feedback, the trained vital sign rehabilitation prediction model is obtained.

[0037] According to one embodiment of the present invention, the loss function of the physical sign rehabilitation prediction model is determined based on the historical age, the historical discharge treatment data, the historical measured physical sign rehabilitation data, and the historical predicted physical sign rehabilitation data, including: determining the loss function LOSS of the physical sign rehabilitation prediction model according to formula (1).

[0038] (1),

[0039] in, This represents the historical measured vital signs and rehabilitation data of the i-th historical patient at the h-th historical home-staying time. This provides the historical predicted vital sign recovery data for the i-th historical patient at the h-th historical home-based time point. Let be the number of days of hospitalization for the i-th historical patient. Standard length of stay, Let i be the total historical discharge treatment cost of the i-th historical patient. The standard total treatment price, For the i-th historical patient, the type of historical chronic disease. For the j-th patient, the type of chronic disease is... Let be the historical age of the i-th historical patient. Let S be the age of the j-th patient, J be the number of patients, S be the number of physiological indicators, and H be the number of historical home-based time periods. Let N be the number of historical patients in the e-th training batch, and N be the number of training batches, where j ≤ J, h ≤ H, and i ≤ 1. , e≤N, and j, a, h, x, i, J, H, Both N and N are positive integers.

[0040] According to an embodiment of the present invention, in formula (1), The relative difference between the historical measured vital signs and rehabilitation data of the i-th historical patient at the h-th historical home time and the historical predicted vital signs and rehabilitation data of the i-th historical patient at the h-th historical home time. This represents the ratio between the historical hospital stay and the standard hospital stay for the i-th historical patient. A larger ratio indicates a longer historical hospital stay, suggesting a more severe condition and slower post-discharge recovery. Therefore, variations in historical hospital stay have a greater impact on the error of historical predictions of recovery. The standard hospital stay can be 14 days. This represents the ratio between the historical total discharge treatment price and the standard total treatment price for the i-th historical patient. A larger ratio indicates a higher historical total discharge treatment price, signifying a more severe condition and slower post-discharge recovery. Therefore, changes in the historical total discharge treatment price have a greater impact on the error of historical predictions of recovery. The standard total treatment price can be 10,000 yuan. This indicates a positive correlation between historical hospital stays and total historical discharge treatment costs and the severity of a patient's condition. For example, higher treatment costs require more medical resources, longer hospital stays necessitate longer observation periods, and generally indicate a more severe condition. Therefore, placing historical hospital stays and total historical discharge treatment costs in the numerator position signifies that the larger these figures are relative to their respective standard values, the more severe the condition. and The larger the value, the greater the impact on the error of historical prediction of physical signs and rehabilitation data. Let the similarity between the historical chronic disease type of the i-th patient and the chronic disease type of the j-th patient be denoted as . Let be the similarity between the historical age of the i-th historical patient and the age of the j-th patient. To achieve similar rehabilitation assessment results, the closer the patient's chronic disease type and age are to those of the historical patients, the better. and The higher the value, the more similar the patient's chronic disease type and age are to the historical chronic disease type and age of historical patients, and the greater its reference value. Therefore, its weight is higher. The weight of the h-th historical home time is used to reasonably weight the errors of different historical home times in the loss function. For the i-th historical patient, the accuracy of the historical predicted physical signs and rehabilitation data of the h-th historical home time output by the physical signs and rehabilitation prediction model is usually higher than that of the historical predicted physical signs and rehabilitation data of the (h+1)-th historical home time. That is, the longer the time interval between a certain historical home time and the first historical home time, the less accurate the prediction result is. To improve training efficiency, its weight is set higher. Conversely, the more accurate the prediction result is, the lower its weight is. The weights for the e-th training batch are used to reasonably weight the relative errors of different training batches in the loss function. For historical patients in the (e+1)-th training batch, the accuracy of the historical predicted physical signs and rehabilitation data of the (e+1)-th training batch output by the physical sign and rehabilitation prediction model is usually higher than that of the historical predicted physical sign and rehabilitation data of the e-th training batch. That is, the shorter the time interval between a training batch and the first training batch, the less accurate the prediction result is. To improve training efficiency, the weights are set higher. Conversely, the more accurate the prediction result is, the lower the weights are. This allows for assigning higher weights to items with lower accuracy, thereby improving training intensity and efficiency.

[0041] According to one embodiment of the present invention, utilizing , , , and The training loss function is obtained by weighted averaging the relative differences of historical predicted vital sign rehabilitation data of multiple patients at multiple historical home times in the e-th training batch. During the training of the vital sign rehabilitation prediction model, some parameters inside the model are adjusted by backpropagating the loss function to reduce its value, thereby improving the accuracy of the vital sign rehabilitation prediction model and obtaining the trained vital sign rehabilitation prediction model.

[0042] In this way, the impact of historical hospital stays and total historical discharge treatment costs on the severity of a patient's condition can be used to determine the influence of these data on the error of historical predicted vital sign rehabilitation data. Based on this influence and the relative difference between historical measured vital sign rehabilitation data and historical predicted vital sign rehabilitation data, weights are assigned based on the patient's chronic disease type and age. Weights are also assigned based on the characteristics that the more similar the patient's historical chronic disease type and age are to historical patient samples, the greater the reference value. Furthermore, weights are assigned based on the characteristics that shorter time intervals from the start of home isolation lead to higher accuracy, and shorter time intervals from the first batch lead to lower accuracy. This weighted summation of the errors output by the vital sign rehabilitation prediction model for multiple historical patients at multiple historical home isolation times in each training batch yields a loss function. This improves the design accuracy and objectivity of the loss function, thereby increasing training efficiency and improving the accuracy of the vital sign rehabilitation prediction model during training.

[0043] According to one embodiment of the present invention, in step S5, the highest rehabilitation coefficient is determined based on the measured vital sign rehabilitation data and the predicted vital sign rehabilitation data.

[0044] Figure 3 A flowchart for calculating the highest recovery coefficient according to an embodiment of the present invention is shown as an example.

[0045] According to an embodiment of the present invention, step S5 includes: step S51, determining the rehabilitation coefficients of multiple patients at multiple home times based on the measured rehabilitation data and the predicted rehabilitation data; step S52, determining the highest rehabilitation coefficient of each patient based on the maximum value of the rehabilitation coefficients of each patient at multiple home times.

[0046] According to one embodiment of the present invention, for each patient, at multiple home-based observation periods, the measured vital sign recovery data and predicted vital sign recovery data are compared and analyzed to calculate the patient's recovery coefficient at each home-based observation period. This recovery coefficient quantitatively assesses the patient's recovery progress at that home-based observation period, and its value reflects the degree of conformity or deviation between the patient's actual recovery status and the expected recovery status. After calculating the recovery coefficients for all patients at all home-based observation periods, the maximum value is further selected from the recovery coefficients of each patient at multiple home-based observation periods. This maximum value is the highest recovery coefficient achieved by the patient during the entire home-based observation period (which can be 3 days or 7 days), representing the patient's optimal state or highest level of recovery progress during that time period.

[0047] According to one embodiment of the present invention, determining the rehabilitation coefficients of multiple patients at multiple home-based times based on the measured vital sign rehabilitation data and the predicted vital sign rehabilitation data includes: determining the rehabilitation coefficient of the j-th patient at the k-th home-based time according to formula (2). ,

[0048] (2),

[0049] in, This represents the measured vital signs and rehabilitation data of the j-th patient at the k-th home time. This is the predicted vital sign recovery data for the j-th patient at the k-th home time, where K is the number of home time, k≤K, and both k and K are positive integers, and if is a conditional function.

[0050] According to an embodiment of the present invention, in formula (2), The conditional function value is given when the measured vital sign recovery data of the j-th patient at the k-th home time is less than or equal to the predicted vital sign recovery data of the j-th patient at the k-th home time. Otherwise, the conditional function value is 1. If the measured vital sign recovery data of the j-th patient at the k-th home time is less than or equal to the predicted vital sign recovery data of the j-th patient at the k-th home time, it means that the actual recovery degree of the j-th patient at the k-th home time is less than or equal to the predicted recovery degree. The relative difference between the predicted vital signs and rehabilitation data of patient j at the k-th home time and the measured vital signs and rehabilitation data of patient j at the k-th home time is the larger the relative difference, the smaller the actual degree of rehabilitation of patient j at the k-th home time. The weight of the kth home-stay time is used to reasonably weight the error of the degree of recovery at different home-stay times. The accuracy of the predicted vital signs and rehabilitation data at the kth home-stay time output by the vital signs and rehabilitation prediction model is usually higher than that at the (k+1)th home-stay time. That is, the shorter the time interval between a certain home-stay time and the first home-stay time, the more accurate the prediction result is, and therefore, the higher its weight. and Multiplying these values ​​yields the recovery coefficient for the j-th patient at the k-th home time. The larger the recovery coefficient, the closer the actual recovery level is to or exceeds the predicted recovery level, and the better the patient's home rehabilitation effect.

[0051] In this way, weights can be set based on the characteristic that the shorter the time interval from the start of home isolation, the higher the accuracy. This allows for a reasonable weighting of the recovery degree error at different home isolation times, resulting in a recovery coefficient that reflects the degree of deviation between the patient's actual recovery status and the expected recovery status, thereby improving the accuracy and reliability of the recovery coefficient.

[0052] According to one embodiment of the present invention, in step S6, patient age information is obtained, and the number of acute exacerbations and hospitalization frequency for each patient in the past year are further counted and recorded. The number of acute exacerbations refers to the number of times a patient required emergency medical intervention due to acute exacerbations or complications of a chronic disease during that period, typically calculated by reviewing the patient's medical records. The hospitalization frequency refers to the number of times a patient was hospitalized for chronic disease-related reasons during that period, which can be obtained by reviewing the patient's hospitalization records.

[0053] According to one embodiment of the present invention, in step S7, a risk coefficient is determined based on the age, the number of acute attacks, and the frequency of hospitalization.

[0054] Figure 4 A flowchart for calculating the risk coefficient according to an embodiment of the present invention is shown as an example.

[0055] According to an embodiment of the present invention, step S7 includes: step S71, obtaining the frequency of family member participation in care through nursing records; step S72, if the frequency of family member participation in care is 3 times or more per week, then determining the patient comfort outcome as 1; step S73, if the frequency of family member participation in care is less than 3 times per week, then determining the patient comfort outcome as 0; step S74, determining a risk coefficient based on the patient comfort outcome, the age, the number of acute attacks, and the frequency of hospitalization.

[0056] According to one embodiment of the present invention, the frequency of family members' involvement in care is statistically recorded in detail by reviewing the patient's nursing records. The nursing records contain relevant information about family members' involvement, such as care time and content. By analyzing this information, the frequency of family members' involvement can be accurately determined, i.e., the number of times they participate in care each week. The patient's comfort outcome is assessed based on the frequency of family members' involvement. The assessment criteria are set as follows: if the frequency of family members' involvement is 3 times or more per week, the patient's comfort outcome is assessed as 1, indicating that the family members' frequent involvement has improved the patient's comfort to some extent; if the frequency of family members' involvement is less than 3 times per week, the patient's comfort outcome is assessed as 0, indicating that the frequency of family members' involvement is relatively low, and the patient's comfort is low. A risk coefficient is determined by comprehensively considering multiple factors, including the patient's comfort outcome, age, number of acute attacks, and frequency of hospitalization. This risk coefficient directly reflects the patient's health risk status.

[0057] According to one embodiment of the present invention, determining a risk coefficient based on the patient comfort outcome, the age, the number of acute attacks, and the frequency of hospitalization includes: determining the risk coefficient of the j-th patient according to formula (3). ,

[0058] (3),

[0059] in, Let j be the age of the j-th patient. Let j be the number of acute attacks for the j-th patient. Let j be the hospitalization frequency of the j-th patient. For patient comfort outcomes, , and The preset weights are used, and max is the function to find the maximum value.

[0060] According to one embodiment of the present invention, in formula (3), This is the ratio between the number of acute attacks in patient j and the maximum number of acute attacks among all patients. The larger this ratio, the more acute attacks patient j has, and the more severe the patient's condition. This is the ratio between the hospitalization frequency of patient j and the maximum hospitalization frequency of all patients. The larger this ratio, the higher the hospitalization frequency of patient j, and the higher the risk of disease exacerbation for that patient. Let be the ratio of the age of the j-th patient to 100. The larger this ratio, the older the patient, the higher the frequency of acute attacks and hospitalizations, i.e., the higher the risk of the patient's condition recurrence. Therefore, the higher the weight of this ratio. The value is calculated as 1 minus the patient's comfort outcome. The larger this difference, i.e., the patient's comfort outcome is 0, the higher the risk of a seizure. , and The risk coefficient can be obtained by weighted summation, where, , and The values ​​can be 0.3, 0.5, and 0.2, respectively. The higher the risk factor, the higher the risk of the patient's condition recurring.

[0061] In this way, multiple factors such as the patient's age, number of acute attacks, frequency of hospitalization, and comfort outcomes can be considered, and weights can be set based on the characteristic that the older the patient, the higher the risk of disease recurrence. By determining the risk coefficient, we can understand the disease risk level of each patient, which helps to develop a more accurate follow-up plan and provide differentiated follow-up services for patients with different risk levels.

[0062] According to one embodiment of the present invention, in step S8, the follow-up order and follow-up frequency are determined based on the highest recovery coefficient and the risk coefficient.

[0063] Figure 5 A flowchart illustrating the arrangement of follow-up sequence and frequency according to an embodiment of the present invention is shown as an example.

[0064] According to an embodiment of the present invention, step S8 includes: step S81, arranging the highest recovery coefficients of multiple patients in ascending order to determine a follow-up sequence; step S82, determining the follow-up order according to the follow-up sequence; step S83, if the risk coefficient is greater than or equal to a preset risk coefficient, determining the follow-up frequency as weekly; step S84, if the risk coefficient is less than the preset risk coefficient, determining the follow-up frequency as monthly.

[0065] According to one embodiment of the present invention, the calculated highest recovery coefficients for each patient are summarized and arranged in ascending order from lowest to highest. Patients with higher highest recovery coefficients indicate relatively better recovery or faster recovery progress, while patients with lower highest recovery coefficients indicate poorer recovery or slower progress. Ascending order provides a visual comparison of patient recovery status, and descending order of the highest recovery coefficients determines the follow-up sequence. This follow-up sequence is a list of patients arranged in ascending order of their highest recovery coefficients, showing the priority of each patient in the follow-up process; for example, the patient with the lowest highest recovery coefficient is listed first. In subsequent follow-up arrangements, follow-ups will be conducted sequentially according to the follow-up sequence, ensuring that patients with poorer recovery or slower progress receive timely and effective attention and intervention, resulting in a rational allocation of medical resources and prioritizing patients with more urgent recovery needs. The risk coefficient for each patient is compared with a preset risk threshold (e.g., 0.6). If a patient's risk coefficient is greater than or equal to a preset risk coefficient, the patient is classified as high-risk, indicating a higher risk of disease flare-ups and requiring closer monitoring and attention. Therefore, the follow-up frequency is set at once a week to promptly grasp changes in the patient's condition and take appropriate treatment measures. Conversely, if a patient's risk coefficient is less than the preset risk coefficient, the patient is classified as low-risk, indicating a relatively low risk of disease flare-ups. The follow-up frequency can be appropriately reduced to once a month, ensuring chronic disease monitoring while improving the efficiency of medical resource utilization.

[0066] The big data-based method for precise follow-up of chronic diseases according to embodiments of the present invention, through discharge diagnosis and treatment data, discharge vital sign data, and home vital sign data, helps to more accurately assess the patient's health status. Utilizing a trained vital sign rehabilitation prediction model, it can predict the patient's predicted vital sign data at multiple home moments, which helps to quantitatively assess the patient's rehabilitation progress and effects. The highest rehabilitation coefficient can more accurately identify the priority of follow-up, and the risk coefficient helps to identify high-risk patients. That is, personalized follow-up plans can be formulated based on the highest rehabilitation coefficient and the risk coefficient, improving the accuracy and effectiveness of follow-up and reducing the waste of medical resources. When determining the loss function of the vital signs rehabilitation prediction model, the impact of historical hospital stays and historical total treatment costs on the severity of the patient's condition can be used to determine the influence of these data on the error of historical vital signs rehabilitation data. Based on this influence and the relative difference between historical measured vital signs rehabilitation data and historical predicted vital signs rehabilitation data, weights are set based on the patient's chronic disease type and age. Weights are also set based on the characteristics that the more similar the patient's sample has to historical chronic disease type and age, the greater the reference value. Furthermore, weights are set based on the characteristics that the shorter the time interval from the start of home isolation, the higher the accuracy, and the shorter the time interval from the first batch, the lower the accuracy. In this way, the errors output by the vital signs rehabilitation prediction model for multiple historical patients at multiple historical home isolation times in each training batch are weighted and summed to obtain the loss function. This improves the design accuracy and objectivity of the loss function, thereby improving training efficiency and the accuracy of the vital signs rehabilitation prediction model during the training process. When determining the recovery coefficient for multiple patients at multiple home-based times, weights can be assigned based on the characteristic that shorter time intervals from the start of home-based care result in higher accuracy. This allows for a reasonable weighting of the recovery degree errors at different home-based times, yielding a recovery coefficient that reflects the deviation between the patient's actual and expected recovery status, thus improving the accuracy and reliability of the recovery coefficient. When determining the risk coefficient, multiple factors can be considered, including the patient's age, number of acute exacerbations, frequency of hospitalizations, and comfort outcomes. Weights can be assigned based on the characteristic that older patients have a higher risk of relapse. Determining the risk coefficient allows for understanding the risk level of each patient, facilitating the development of more precise follow-up plans and providing differentiated follow-up services for patients with different risk levels.

[0067] Figure 6An exemplary block diagram of a big data-based precision follow-up system for chronic diseases according to an embodiment of the present invention is shown. The system includes: a discharge diagnosis and treatment data and discharge vital sign data module, used to acquire discharge diagnosis and treatment data and discharge vital sign data for multiple patients, wherein the discharge diagnosis and treatment data includes total discharge treatment cost, length of hospital stay, and chronic disease type, and the discharge vital sign data includes discharge weight and multiple discharge physiological indicators; a home vital sign data module, used to acquire home vital sign data for multiple patients at multiple home times, wherein the home vital sign data includes home weight and multiple home physiological indicators; and a measured vital sign rehabilitation data module, used to determine measured vital sign rehabilitation based on the home vital sign data, standard weight, and multiple standard physiological indicators. The system includes: a data prediction module for predicting rehabilitation data; a data prediction model for inputting the discharge diagnosis and treatment data and the discharge vital signs data into a trained model to obtain predicted rehabilitation data for multiple patients at multiple home times; a maximum rehabilitation coefficient module for determining the maximum rehabilitation coefficient based on the measured rehabilitation data and the predicted rehabilitation data; an age, number of acute attacks, and hospitalization frequency module for obtaining the age of multiple patients, as well as the number of acute attacks and hospitalization frequency within one year; a risk coefficient module for determining the risk coefficient based on the age, the number of acute attacks, and the hospitalization frequency; and a follow-up order and follow-up frequency module for determining the follow-up order and follow-up frequency based on the maximum rehabilitation coefficient and the risk coefficient.

[0068] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0069] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The objectives of the present invention have been fully and effectively achieved. The functions and structural principles of the present invention have been demonstrated and explained in the embodiments, and any variations or modifications may be made to the implementation of the present invention without departing from the stated principles.

Claims

1. A method for precise follow-up of chronic diseases based on big data, characterized in that, include: The process involves acquiring discharge diagnosis and treatment data and discharge vital sign data for multiple patients. The discharge diagnosis and treatment data includes total treatment cost, length of hospital stay, and type of chronic disease. The discharge vital sign data includes discharge weight and multiple discharge physiological indicators. At multiple home-based times, home-based vital sign data for multiple patients is acquired. This home-based vital sign data includes home weight and multiple home-based physiological indicators. Based on the home-based vital sign data, standard weight, and multiple standard physiological indicators, measured vital sign rehabilitation data is determined. The discharge diagnosis and treatment data and discharge vital sign data are input into a trained vital sign rehabilitation prediction model to obtain predicted vital sign rehabilitation data for multiple patients at multiple home-based times. Based on the measured vital sign rehabilitation data and the predicted vital sign rehabilitation data, the highest rehabilitation coefficient is determined. The ages of multiple patients, as well as the number of acute exacerbations and hospitalization frequency within one year, are acquired. Based on the age, the number of acute exacerbations, and the hospitalization frequency, a risk coefficient is determined. Based on the highest rehabilitation coefficient and the risk coefficient, the follow-up order and follow-up frequency are determined. The highest recovery coefficient is determined based on the measured vital sign recovery data and the predicted vital sign recovery data, including: determining the recovery coefficients of multiple patients at multiple home time periods based on the measured vital sign recovery data and the predicted vital sign recovery data; and determining the highest recovery coefficient for each patient based on the maximum value of the recovery coefficient for each patient at multiple home time periods. Based on the measured vital sign recovery data and the predicted vital sign recovery data, the recovery coefficients of multiple patients at multiple home-based times are determined, including: according to the formula Determine the recovery coefficient of the j-th patient at the k-th home time. ,in, This represents the measured vital signs and rehabilitation data of the j-th patient at the k-th home time. This is the predicted vital sign recovery data for the j-th patient at the k-th home time, where K is the number of home time, k≤K, and both k and K are positive integers, and if is a conditional function.

2. The method for precise follow-up of chronic diseases based on big data according to claim 1, characterized in that, Based on the home-based vital sign data, standard weight, and multiple standard physiological indicators, the measured vital sign rehabilitation data are determined, including: obtaining home-based vital sign vectors for multiple patients at multiple home-based times based on home-based vital sign data for multiple patients at multiple home-based times; obtaining standard vital sign vectors based on the standard weight and the standard physiological indicators; and determining the similarity between the home-based vital sign vectors and the standard vital sign vectors as the measured vital sign rehabilitation data for the j-th patient at the k-th home-based time, where k and j are both positive integers.

3. The method for precise follow-up of chronic diseases based on big data according to claim 1, characterized in that, The training steps of the vital sign rehabilitation prediction model include: acquiring historical discharge medical data and historical discharge vital sign data of multiple historical patients, wherein the historical discharge medical data includes historical total treatment cost, historical length of hospital stay, and historical chronic disease type, and the historical discharge vital sign data includes historical discharge weight and multiple historical discharge physiological indicators; acquiring historical home vital sign data of historical patients at multiple historical home times, wherein the historical home vital sign data includes historical home weight and multiple historical home physiological indicators; and based on the historical home vital sign data, standard weight, and standard physiological indicators... The process involves: determining historical measured vital sign rehabilitation data; processing the historical discharge treatment data and historical discharge vital sign data using a vital sign rehabilitation prediction model to obtain historical predicted vital sign rehabilitation data for multiple patients at multiple historical home-based times; obtaining the historical age of the historical patients; determining the loss function of the vital sign rehabilitation prediction model based on the historical age, the historical discharge treatment data, the historical measured vital sign rehabilitation data, and the historical predicted vital sign rehabilitation data; and training the vital sign rehabilitation prediction model based on the loss function to obtain the trained vital sign rehabilitation prediction model.

4. The method for precise follow-up of chronic diseases based on big data according to claim 3, characterized in that, Based on the historical age, the historical discharge and treatment data, the historical measured vital sign rehabilitation data, and the historical predicted vital sign rehabilitation data, the loss function of the vital sign rehabilitation prediction model is determined, including: according to the formula Determine the loss function LOSS of the vital sign rehabilitation prediction model, where, This represents the historical measured vital signs and rehabilitation data of the i-th historical patient at the h-th historical home-staying time. This provides the historical predicted vital sign recovery data for the i-th historical patient at the h-th historical home-based time point. Let be the number of days of hospitalization for the i-th historical patient. Standard length of stay, Let i be the total historical discharge treatment cost of the i-th historical patient. The standard total treatment price, For the i-th historical patient, the type of historical chronic disease. For the j-th patient, the type of chronic disease is... Let be the historical age of the i-th historical patient. Let S be the age of the j-th patient, J be the number of patients, S be the number of physiological indicators, and H be the number of historical home-based time periods. Let N be the number of historical patients in the e-th training batch, and N be the number of training batches, where j ≤ J, h ≤ H, and i ≤ 1. , e≤N, and j, a, h, x, i, J, H, Both N and N are positive integers.

5. The method for precise follow-up of chronic diseases based on big data according to claim 1, characterized in that, The risk factor is determined based on the patient's age, the number of acute attacks, and the frequency of hospitalization, including: obtaining the frequency of family member involvement in care through nursing records; if the frequency of family member involvement is 3 times or more per week, the patient's comfort outcome is determined to be 1; if the frequency of family member involvement is less than 3 times per week, the patient's comfort outcome is determined to be 0; and the risk factor is determined based on the patient's comfort outcome, the patient's age, the number of acute attacks, and the frequency of hospitalization.

6. The method for precise follow-up of chronic diseases based on big data according to claim 5, characterized in that, The risk factor is determined based on the patient's comfort outcome, age, number of acute exacerbations, and frequency of hospitalizations, including: according to the formula Determine the risk coefficient for the j-th patient. ,in, Let j be the age of the j-th patient. Let j be the number of acute attacks for the j-th patient. Let j be the hospitalization frequency of the j-th patient. For patient comfort, , and The preset weights are used, and max is the function to find the maximum value.

7. The method for precise follow-up of chronic diseases based on big data according to claim 1, characterized in that, The follow-up sequence and frequency are determined based on the highest recovery coefficient and the risk coefficient, including: arranging the highest recovery coefficients of multiple patients in ascending order to determine the follow-up sequence; determining the follow-up order based on the follow-up sequence; if the risk coefficient is greater than or equal to a preset risk coefficient, the follow-up frequency is determined to be weekly; if the risk coefficient is less than the preset risk coefficient, the follow-up frequency is determined to be monthly.

8. A big data-based system for precise follow-up of chronic diseases, used to execute the big data-based method for precise follow-up of chronic diseases as described in any one of claims 1-7, characterized in that, include: The hospital discharge diagnosis and treatment data and discharge vital sign data module is used to acquire discharge diagnosis and treatment data and discharge vital sign data for multiple patients. The discharge diagnosis and treatment data includes the total treatment cost at discharge, length of hospital stay, and type of chronic disease. The discharge vital sign data includes discharge weight and multiple discharge physiological indicators. The home vital sign data module is used to acquire home vital sign data for multiple patients at multiple home times. The home vital sign data includes home weight and multiple home physiological indicators. The measured vital sign rehabilitation data module is used to determine measured vital sign rehabilitation data based on the home vital sign data, standard weight, and multiple standard physiological indicators. The predicted vital sign rehabilitation data module is used to... Discharge diagnosis and treatment data and discharge vital sign data are input into a trained vital sign rehabilitation prediction model to obtain predicted vital sign rehabilitation data for multiple patients at multiple home times; a maximum rehabilitation coefficient module is used to determine the maximum rehabilitation coefficient based on the measured vital sign rehabilitation data and the predicted vital sign rehabilitation data; an age, number of acute attacks, and hospitalization frequency module is used to obtain the age of multiple patients, as well as the number of acute attacks and hospitalization frequency within one year; a risk coefficient module is used to determine the risk coefficient based on the age, the number of acute attacks, and the hospitalization frequency; and a follow-up order and follow-up frequency module is used to determine the follow-up order and follow-up frequency based on the maximum rehabilitation coefficient and the risk coefficient.

Images