Disclosure of Invention
The invention provides a medical guidance method, a medical guidance device and a server based on clinical data, which are used for overcoming the defects of the prior art, and aims to provide medical guidance for a patient according to the clinical data, establish a quick, sensitive and reliable information bridge between the patient and excellent doctors and departments, simplify unnecessary links and improve efficiency.
In order to achieve the above object, the present invention provides a medical guidance method based on clinical data, comprising the following steps:
acquiring clinical data of all disease patients;
establishing an index database related to a certain disease patient by limiting keywords;
establishing a data index for measuring the effectiveness of a doctor for treating the certain disease;
extracting ID, doctor, department, vital sign, diagnosis and examination information of each patient from the clinical data according to the index database;
constructing a weighted heterogeneous star data model according to the extracted information and the data indexes;
and ranking the weighted heterogeneous star-shaped data model to obtain ranking information suitable for the treatment level of doctors of the patients with the certain disease.
Preferably, after the acquiring clinical data of all disease patients in the clinical database, the method further comprises:
and cleaning the clinical data.
Preferably, the cleansing the clinical data comprises:
removing the recorded data of the patient when the data missing rate of the recorded data of the patient is greater than a preset threshold value;
when the data missing rate of the recorded data of the patient is smaller than a preset threshold value, the recorded data of the patient are supplemented;
selecting a record with the closest time for clinical data of each patient;
and dispersing the value of the test result of the patient to a preset interval.
Preferably, the data indicators include:
the examination information of the patient at different stages and the outcome type of the disease.
Preferably, the constructing a weighted heterogeneous star data model according to the relational database and the data indexes includes:
establishing a heterogeneous star-shaped data model which takes the ID information of the patient as a main key and takes the doctor, the department, the vital sign, the diagnosis and the inspection information of the patient as attribute objects according to the extracted information;
and calculating the weight of each attribute object of the heterogeneous star data model according to the data index.
Preferably, the ranking the weighted heterogeneous star-shaped data model to obtain ranking information of the doctor treatment level of the certain disease patient includes:
performing iterative operation by using a MedRank ranking algorithm through a first formula until a result converges to a stable value, and ranking the weighted heterogeneous star-shaped data model so as to obtain ranking information of the doctor treatment level of the patient with the certain disease;
wherein, the first formula is:
wherein t is equal to {1, …, n-1}, n is a positive integer greater than 1, and X is
1Is a target type, representing drug information, X
tThe t-th object type, C the center type, representing the patient,
is X
1The ranking score of the object type at the current iteration is U is | X
1|×|X
1Identity matrix of |, X |
1I is X
1The total number of type objects, alpha, is determined as U/| X
1Weight of the | term, W
ABA weighted adjacency matrix, which is the object A and B, represents the weighted linkage between the two,
for normalizing the diagonal matrix of rows, the diagonal value of the ith row is W
ABThe sum of the ith row.
In order to achieve the above object, the present invention further provides a medical advice guidance apparatus based on clinical data, including:
the data acquisition module is used for acquiring clinical data of all disease patients;
an index data module for establishing an index database related to a patient with a certain disease by limiting keywords;
the measuring index module is used for establishing a data index for measuring the effectiveness of a doctor for treating the certain disease;
the information extraction module is used for extracting the ID, doctor, department, vital sign, diagnosis and inspection information of each patient from the clinical data according to the index database;
the model module is used for constructing a weighted heterogeneous star data model according to the extracted information and the data indexes;
and the ranking module is used for ranking the weighted heterogeneous star-shaped data model and acquiring ranking information of the doctor treatment level of the certain disease.
Preferably, the clinical data based medication recommendation device further comprises:
the data cleaning module is used for cleaning the clinical data;
the model module comprises:
the construction module is used for establishing a heterogeneous star-shaped data model which takes the ID information of the patient as a main key and takes the doctor, the department, the vital sign, the diagnosis and the inspection information of the patient as attribute objects according to the extracted information;
and the weight module is used for calculating the weight of each attribute object of the heterogeneous star data model according to the data indexes.
Preferably, the ranking module is configured to perform iterative operation by using a MedRank ranking algorithm through a first formula until a result converges to a stable value, so as to rank the weighted heterogeneous star-shaped data model, and thus obtain ranking information of a doctor treatment level of the patient with the certain disease;
wherein, the first formula is:
wherein t is equal to {1, …, n-1}, n is a positive integer greater than 1, and X is
1Is a target type, representing drug information, X
tThe t-th object type, C the center type, representing the patient,
is X
1The ranking score of the object type at the current iteration is U is | X
1|×|X
1Identity matrix of |, X |
1I is X
1The total number of type objects, alpha, is determined as U/| X
1Weight of the | term, W
ABA weighted adjacency matrix, which is the object A and B, represents the weighted linkage between the two,
diagonal matrix for normalizing rowsWherein the diagonal value of the ith row is W
ABThe sum of the ith row.
In order to achieve the above object, the present invention further provides a server, which includes the above medical guidance device based on clinical data.
According to the hospitalizing guidance method, the hospitalizing guidance device and the server based on clinical data, provided by the invention, the clinical data of all disease patients are collected, then an index database is established by limiting keywords related to a certain disease, and then a data index for measuring the effectiveness of a doctor in treating the disease is established; extracting relevant information of the patient from the clinical database according to the mapping relation between the index database and the clinical database; constructing a data model according to the extracted information and the data indexes; finally, analyzing, processing and ranking the data model by a mathematical method to finally obtain the ranking information of doctors treating the level of the disease patients; the method and the system realize the auxiliary decision for the patient to select the doctor according to the clinical data, thereby establishing a quick, sensitive and reliable information bridge between the patient and excellent doctors and departments, simplifying unnecessary links, shortening the hospitalizing time and improving the hospitalizing efficiency.
Example one
As shown in fig. 1, the present invention provides a medical guidance method based on clinical data, comprising the following steps:
step S10, acquiring clinical data of all disease patients; the clinical data here is derived from existing data, and may be clinical data of one or more hospitals. The information about each patient of these clinical data is in a uniform format. The method mainly comprises the following steps: patient ID (identification number), diagnosis of disease, pre-medication test results, vital signs, doctor, department, post-medication test results, outcome status, etc.
Step S20, establishing an index database related to a disease patient by limiting keywords; the keyword may be a word related to the name of a disease, and is directed to the search of information for diagnosing the disease, such as hypertension, heart disease, coronary heart disease, etc. The index database contains links of clinical data of all patients related to the keywords, the clinical data of all patients related to the keywords or the medical subject vocabulary can be found through mapping relations in the index database, and information contained in the index database is from content in the clinical database.
The clinical data in this example is acquired from the medical information system of the central university, xiangya three hospitals over 5 days, and the information of the doctor of the hypertensive who has two blood pressure measurements and vital signs and test results before and after the administration. The data includes the basic information and examination indexes such as ID number representing the patient, doctor, department, basic information, blood pressure before and after the vital signs are taken, treatment medicine, prognosis condition, age, sex, smoking, blood fat, blood sugar, BMI, EGFR, CRP, creatinine, microalbuminuria and the like, and diagnoses such as heart failure, coronary heart disease, diabetes and the like, wherein the basic information and the examination indexes do not include the privacy information of the patient.
The therapeutic drugs for hypertension patients mainly include diuretics, β receptor blocker, α receptor blocker, Angiotensin Converting Enzyme Inhibitor (ACEI), Calcium Channel Blocker (CCB), angiotensin receptor Antagonist (ARB), vasodilator, ganglion blocker, single drug or combination drug.
Clinical data of all hypertension patients in a hospital information system are analyzed, and clinical data information of patients including patient ID, diastolic pressure/systolic pressure before and after medication, main doctors, departments, antihypertensive drugs and the like is acquired.
Step S30, establishing a data index for measuring the effectiveness of a doctor in treating the certain disease; the data indexes comprise inspection information of different stages of the patient and the classification type of the disease condition, taking the patient with the hypertension as an example: 1) the diastolic pressure/systolic pressure is obviously reduced during the treatment period of the antihypertensive drug; 2) the condition of the patient is improved when the patient is discharged. In the invention, the blood pressure reduction condition before and after the administration of the drug of the hypertensive during one visit and the state of illness are used together to measure the effectiveness of the blood pressure reduction scheme or the blood pressure reduction drug during the visit, and a threshold value can be set, so that whether the blood pressure reduction of the hypertensive before and after the administration of the drug during one visit reaches the threshold value or not can be determined; the improvement can be set to several grade types, such as mild, moderate, good, healing, etc.
Step S40, extracting ID, doctor, department, vital sign, diagnosis and inspection information of each patient from the clinical data according to the index database; if the information in the clinical data is not complete, the clinical data may be cleaned after step S10 to remove the attribute with the excessive missing data rate and the patient record, and take data supplementary measures for the attribute with the small missing data rate.
Step S50, constructing a weighted heterogeneous star data model according to the extracted information and the data indexes; the method comprises the following steps:
step S51, according to the extracted information, establishing a star-shaped heterogeneous data model which takes the ID information of the patient as a main key and takes the doctor, the department, the vital sign, the diagnosis and the inspection information of the patient as attribute objects; see fig. 2;
and step S52, calculating the weight of each attribute object of the heterogeneous star data model according to the data index.
And constructing a relational data structure taking the patient as a central object and taking characteristics such as doctor, department, age, diastolic pressure before and after medication, systolic pressure, heart rate, hemoglobin and the like as attribute objects aiming at the clinical data of each patient, and calculating the weight of the attribute objects of each relational data structure according to data indexes to form a plurality of weighted heterogeneous star-shaped data models.
Wherein the vital signs include age, systolic pressure/diastolic pressure before administration, and the test results include test values of EGFR, urea, and hemoglobin. The weight of the attribute object of the star-shaped heterogeneous data model edge is calculated on the basis of the following two factors:
the blood pressure of the patient is reduced: if the patient had a pre-medication diastolic pressure <90 and systolic pressure < 140:
a) post-dose systolic pressure <140 and diastolic pressure <90, weight ═ 0;
b) the post-medication systolic pressure > 140 or diastolic pressure > 90, weight-1;
if pre-medication diastolic pressure > 90 or systolic pressure > 140:
a) post-dose systolic pressure <140 and diastolic pressure <90, weight ═ 1;
b) systolic blood pressure reduction >20 or diastolic blood pressure reduction >10, weight 0.5;
c) systolic blood pressure drop <20 and diastolic blood pressure drop <10, weight ═ 0;
d) the systolic pressure or diastolic pressure rises, weight is-1;
the patients' condition is classified as follows:
if the patient's condition is improved, weight is 1;
if the patient's condition is not cured, the weight is-0.5;
if the patient dies, weight-1;
if the patient's state of illness belongs to other conditions, weight is 0;
finally, other index weights of the patient are as follows:
weight _ transfer + a weight _ blood pressure decrease condition (1-a)
Doctor's weight: the average of all patients' weights treated by the doctor;
weight of department: average of all patient weights treated by the department.
And step S60, ranking the weighted heterogeneous star-shaped data model to obtain ranking information of the doctor treatment level of the disease patient.
And ranking the weighted heterogeneous star-shaped data model constructed in the step S50 by adopting a MedRank ranking algorithm to obtain the ranking of the hypertension disease treatment level of each doctor. The MedRank algorithm carries out iterative operation on the weighted heterogeneous star data model constructed by each patient based on the following algorithm until the result converges to a stable quality:
X
tis at the t
thObject type, X
1For the target type, referred to as the hypotensive regime in this study, C is the central type, referred to as patient in this example, W
ABA weighted adjacency matrix, which is the object A and B, represents the weighted linkage between the two,
for normalizing the diagonal matrix of rows, the diagonal value of the ith row is W
ABThe sum of the ith row. Table 1 below is a ranking of the clinical data for the level of treatment of all hypertensive patients by the attending physician:
TABLE 1
According to the scheme of the invention, through the analysis of clinical data, the ranking condition of treatment levels of doctors and departments related to various diseases can be calculated, the diagnosis and treatment levels of the departments and the doctors can be evaluated, the doctor-seeing guidance is provided for patients during the doctor-seeing, the patients are helped to scientifically and reasonably select the departments and the doctors to see the doctor, and the doctor-seeing time is shortened.
In an embodiment of the present invention, referring to fig. 3, after step S10, the method further includes:
step S11, cleaning the clinical data; take hypertension as an example: the clear clinical data includes the blood pressure before and after the medicine, doctors, departments and antihypertensive medicines.
Step S11 includes:
step S111, removing the recorded data of the patient when the data loss rate of the recorded data of the patient is larger than a preset threshold value; when the data missing rate of the recorded data of the patient is smaller than a preset threshold value, the recorded data of the patient are supplemented; if necessary data information in the clinical data is missing, judging whether the missing rate is greater than a threshold value, and if so, removing the record of the patient; if the value is smaller than the value, the numerical type attribute is filled by using the median (average value) of the attribute value, and the label type attribute is filled by using a mode of randomly generating a label value;
step S112, selecting a record with the most recent time in the clinical data of each patient;
in step S113, the numerical value of the test result of the patient is dispersed to a preset interval.
The cleaned data is more complete, obvious distortion or error data is removed, and the data with universality is adopted, so that the calculation accuracy can be improved, and errors caused by the original data are reduced.
The embodiment of the invention also provides a medical treatment guiding device based on clinical data, which comprises:
the data acquisition module 10 is used for acquiring clinical data of all disease patients;
an index data module 20 for establishing an index database related to a patient of a certain disease by defining keywords;
a measurement index module 30 for establishing a data index for measuring the effectiveness of the doctor in treating the certain disease;
an information extraction module 40, configured to extract ID, doctor, department, vital sign, diagnosis, and examination information of each patient from the clinical data according to the index database;
a model module 50, configured to construct a weighted heterogeneous star data model according to the extracted information and the data index;
and the ranking module 60 is used for ranking the weighted heterogeneous star-shaped data model and acquiring ranking information of the doctor treatment level of the patient with the certain disease.
The clinical data based medication recommendation device further comprises:
a data cleaning module 11, configured to clean the clinical data;
the model module 50 includes:
the construction module 51 is configured to establish a star-shaped heterogeneous data model using the ID information of the patient as a main key and using the doctor, department, vital sign, diagnosis and examination information of the patient as attribute objects according to the extracted information;
and a weight module 52, configured to calculate a weight of each attribute object of the star-shaped heterogeneous data model according to the data index.
The ranking module 60 is configured to perform iterative operation by using a MedRank ranking algorithm through a first formula until a result converges on a stable value, so as to rank the weighted heterogeneous star-shaped data model, thereby obtaining ranking information of a doctor treatment level of a certain disease patient;
wherein, the first formula is:
wherein t is equal to {1, …, n-1}, n is a positive integer greater than 1, and X is
1Is a target type, representing drug information, X
tThe t-th object type, C the center type, representing the patient,
is X
1The ranking score of the object type at the current iteration is U is | X
1|×|X
1Identity matrix of |, X |
1I is X
1The total number of type objects, alpha, is determined as U/| X
1Of the itemWeight, W
ABA weighted adjacency matrix, which is the object A and B, represents the weighted linkage between the two,
for normalizing the diagonal matrix of rows, the diagonal value of the ith row is W
ABThe sum of the ith row.
The data acquisition module 10 of the device acquires the data related to the hypertension patient, including the age, the doctor, the diastolic/systolic pressure before and after medication, the hypotensive drugs and the like, with medication time of more than 5 days according to all the data of the patient in the information of the hospital patient for seeing a doctor, and two times of blood pressure measurement and vital signs and test results before and after medication. The data cleaning index establishing module is used for establishing indexes related to the effectiveness of clinical data, wherein the collected patient records must have attributes such as doctors, departments, antihypertensive treatment, basic information (age, sex and BMI), vital signs (blood pressure and heart rate before medication), test results, diagnosis and outcome transfer; defining a reasonable range of attribute values when cleaning data; for missing data, deleting records or supplementing missing values; constructing a weighted heterogeneous star network model according to the collected clinical data; the MedRank ranking algorithm is adopted to rank the weighted heterogeneous star network model to obtain ranking information of the treatment level of the doctor, and the patient or the family can select the main doctor according to the ranking, so that the recommendation of the hypertension treating doctor is realized.
The embodiment of the invention also provides a server which comprises any one of the clinical data-based medical guidance devices in the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.