TW201837745A - Cross-platform analyzing and display system of clinical data - Google Patents

Abstract

A cross-platform analyzing and display system of clinical data is provided in the present disclosure. The cross-platform analyzing and display system of clinical data includes a major processing module, an automation statistical module and a verification module. The major processing module collects and generates a plurality of clinical medical data, which includes a clinical data collection module and an automated acceleration module. The automated statistical module performs a statistical analysis on the data of the major processing module using a plurality of regression diagnosis methods to generate an analysis result. The verification module verifies a disease matrix and the analysis result of the automated statistical module according to at least a plurality of paper of the clinical data corresponding to a first clinical medical database.

Description

 Cross-platform clinical medical data analysis and display system  

The invention relates to a clinical medical data analysis and display system, and in particular to a clinical medical data analysis and display system capable of accelerating the calculation of massive data.

Comorbidity refers to two or more diseases that are simultaneous or non-simultaneous, and the association between comorbidities is very important in the nosology of clinical medicine. The comorbidity shows a temporary relationship between co-existing disease, and chronic diseases of different ages often affect the establishment of disease classification and clinical decision-making. Complex chronic medical history is closely related to the prognosis of patients (BM J 334, 1016-1017 (2007) and Annals of family medicine 4, 417-422 (2006)), particularly in the elderly medicine and cancer (Jama 291, 2441-2447 (2004)). . As the aging population increases year by year, research topics for comorbidities should be strained, but there is currently insufficient empirical medical evidence to provide relevant diagnosis and treatment decisions (JAMA 294, 716-724 (2005) and Lancet 367, 550-551 (2006) ).

Common causes can be divided into: (1), causal, that is, two or more diseases have a common disease physiology. (2) Complexity, which is related to specific death between diseases. According to the time of occurrence, it can be divided into concurrent, concurrent (intercurrent) and successive comorbidity (Journal of child psychology and psychiatry, and allied disciplines 40, 57-87 (1999)), coexistence. That is, two unrelated diseases exist simultaneously, and intercurrent means that the interaction between comorbidities is affected by the acute phase of the disease, and is usually limited by time.

Medical research on comorbidity has developed dramatically in the last 10 years (Cell March 18, 2011 vol. 144 no. 6 986-998). In the past, studies have used the relationship between 161 diseases and genes provided by a single medical center to provide 1.5 million medical records (PNAS July 10, 2007 vol. 104 no. 28 11694-11699), modelling and calculating phenotypes. The time course and the probability of disease, and the Harvard scholars use a database of 32 million patients (PLoS Comput Biol 5(4): e1000353), which counts patients older than 65 years old. Past medical history, cross-sectional studies, and calculate the relative risk of disease included in the ICD9 diagnostic code. The above research is of great significance for the analysis of comorbidity and clinical medical big data, and proposes the concept of physical and chemical analysis combined with automated analysis of human biological database.

Although the above innovative research has contributed academically, the results still have many obstacles to be solved in the experimental clinical use. Here are the following reasons: First, the database used is low representative and non-recorded. The public's general medical habits; Second, the analytical methods used are cross-cutting research and self-created methods similar to the generational tracking. The evidence and causal derivation and intrinsic validity are lower than the traditional generation tracking research. Third, the use of horizontal Broken research does not calculate important clinical data such as complete disease cycle and mortality; fourth, its automated analysis lacks high-confidence verification, both internal verification in the database and external verification outside the database. Not enough; V. Lack of a rational collection of diagnostic codes for statistic diseases; 6. Lack of elimination of drugs and complications caused by surgery.

In the past, Taiwanese scholars used the information provided by the National Health Insurance Administration Ministry of Health and Welfare to refer to foreign pilot research and establish big data for automated analysis (transverse), but traverse The cross-sectional study and the time-long longitudinal study can vary by tens of times in efficiency, and there are obstacles in computational efficiency.

Therefore, providing an analytical system that can accurately predict the likelihood of a user's disease development is an important issue in the industry today.

In view of this, an embodiment of the present invention provides a cross-platform clinical medical data analysis and display system, which is disposed on a server, wherein the server includes a central control device, a storage device, and a communication device, the cross-platform medical data. The analysis and display system comprises: a core computing main module, an automated statistical module and a verification module. The core computing main module collects and generates a plurality of clinical medical data, including: a clinical medical data collection module, and the disease diagnosis code, the judgment of various dates, and the analysis according to the plurality of data of the at least one first clinical medical database; Classifying and collecting medical data, and converting a plurality of data of the first clinical medical database into a plurality of predetermined format data; and an automated acceleration module according to a multiple nested longitudinal statistical method and the plurality of predetermined Format data, conduct disease relationship analysis, and generate a disease matrix, wherein the disease matrix is an MxN matrix, and establish a large Trajectory database that is different from the original data, and then the operation is performed unless the original database is replaced. Use the same Trajectory database. The automated statistical module utilizes a plurality of regression diagnostic methods to perform statistical analysis on the data of the core computing main module to generate an analysis result; and the verification module, at least according to the clinical clinics corresponding to the first clinical medical database The multiple papers of the data data verify the disease matrix of the core computing main module and a plurality of clinical medical data.

In summary, the cross-platform clinical medical data analysis and display system of the present invention performs further analysis and statistics through the data of a clinical medical database, which greatly reduces the calculation amount, speeds up the processing, and can not reduce the accuracy thereof. In addition, the professional or general version of the analysis results can be obtained from the webpage or application provided by the system, which not only provides users with clear and rapid medical analysis information, but also instantly understands the possible development direction of their health. .

The above described features and advantages of the present invention will be more apparent from the following description.

1‧‧‧Server

10‧‧‧Central control unit

11‧‧‧Storage device

12‧‧‧ display device

13‧‧‧Database

14‧‧‧Communication device

121‧‧‧Computer version of web page display module

122‧‧‧Mobile device application display module

A‧‧‧ core computing main module

A1‧‧‧ Clinical Medical Data Collection Module

A2‧‧‧Automation Acceleration Module

B‧‧‧Automatic Statistics Module

C‧‧‧ verification module

A000-A063, B001-B020, C000-1, C000-2, C001-C010, D001-D028, E001-E012, E014-E016‧‧

F001-F016, G001-G011, H001-H023‧‧‧ blocks

FIG. 1 is a schematic diagram of a server according to an embodiment of the present invention.

2 is a schematic diagram of a cross-platform clinical medical data analysis and display system according to an embodiment of the present invention.

FIG. 3 is a schematic diagram of a display device according to an embodiment of the invention.

FIG. 4 is a schematic diagram of a core computing main module according to an embodiment of the present invention.

FIG. 5 is a schematic diagram of an automated acceleration module according to an embodiment of the present invention.

FIG. 6 is a schematic diagram of a verification module according to an embodiment of the present invention.

FIG. 7 is a schematic diagram of an automatic regression diagnosis sub-module of an automated statistical module according to an embodiment of the present invention.

8 is a flow chart of a server in accordance with an embodiment of the present invention.

FIG. 9 is a schematic diagram of a professional version webpage according to an embodiment of the present invention.

FIG. 10 is a schematic diagram of a normal version webpage according to an embodiment of the present invention.

FIG. 11 is a schematic diagram of an application according to an embodiment of the present invention.

Various illustrative embodiments are described more fully hereinafter with reference to the accompanying drawings. However, the inventive concept may be embodied in many different forms and should not be construed as being limited to the illustrative embodiments set forth herein. Rather, these exemplary embodiments are provided so that this invention will be in the In the drawings, the size and relative sizes of layers and regions may be exaggerated for clarity. Similar numbers always indicate similar components.

It will be understood that, although the terms first, second, third, etc. may be used herein to describe various elements, such elements are not limited by the terms. These terms are used to distinguish one element from another. Thus, a first element discussed below could be termed a second element without departing from the teachings of the inventive concept. As used herein, the term "and/or" includes any of the associated listed items and all combinations of one or more.

The cross-platform clinical medical data analysis and display system will be described below in conjunction with at least one embodiment. However, the following embodiments are not intended to limit the disclosure.

[Embodiment of the cross-platform clinical medical data analysis and display system of the present invention]

1 to FIG. 3, FIG. 1 is a schematic diagram of a server according to an embodiment of the present invention.

2 is a schematic diagram of a cross-platform clinical medical data analysis and display system according to an embodiment of the present invention. FIG. 3 is a schematic diagram of a display device according to an embodiment of the invention.

The embodiment of the invention provides a cross-platform clinical medical data analysis and display system, which is executed on a server 1 in the end of the embodiment. In this embodiment, the server 1 can be disposed at the local or remote end, which is not limited in the present invention.

The server 1 includes a central control device 10, a storage device 11, a display device 12, a database 13, and a communication device 14.

The central control device 10 is electrically connected to the storage device 11 , the display device 12 , the database 13 , and the communication device 14 . In other embodiments, the central control device 10, the storage device 11, the display device 12, the database 13 and the communication device 14, etc., may be a subsystem in the virtual system, which is not limited in the present invention.

The central control device 10 of the server 1 is used to process various data of the cross-platform clinical medical data analysis and display system, and can perform user identification, receive parameters input from the front-end interface, and transmit the operation result to the front-end interface.

The storage device 11 is configured to store various materials of the cross-platform clinical medical data analysis and display system. In the present embodiment, the data library 13 is disposed in the storage device 11.

In the present embodiment, the display device 12 is used to display various display materials of the cross-platform clinical medical data analysis and display system, such as a computer version of a web page and a mobile device version of a web page or a mobile device version application. In this embodiment, the cross-platform clinical medical data analysis and display system can transmit the related display data to the user through the display device 12, so that the user can use the present invention on the computer or on different mobile devices. Cross-platform clinical medical data analysis and display system.

In the present embodiment, the central control device 11 transmits to the user's computer or mobile device via the communication device 14 various display materials of the cross-platform clinical medical data analysis and display system of the present invention. In this embodiment, the communication device 14 can be a wired communication device or a wireless communication device, which is not limited in the present invention.

In this embodiment, the cross-platform clinical medical data analysis and display system includes a core computing main module A, an automated statistical module B, and a verification module C.

The core operation main module A includes a clinical medical data collection module A1 and an automatic acceleration module A2.

The clinical medical data collection module A1 is used to collect various clinical medical materials and convert them into a unified format, classify disease occurrence events, disease types, classify surviving patients, classify experimental group patients, and control group patients.

The clinical medical data collection module A1 includes the steps of cutting date data and disease diagnosis code data from the original clinical medical data in the step of collecting clinical medical data. The Automated Acceleration Module A2 uses statistical principles to automatically combine all diseases until the end of all disease combinations.

The automated statistical module B mainly includes an automated regression diagnostic sub-module. The automated regression diagnostic sub-module is based on the input conditions and uses the regression statistical method for regression diagnosis. The automated statistical sub-module counts the data of a certain disease combination in the system according to the input conditions.

The verification module C further includes an expert verification sub-module for verifying the statistical result in the verification module by the medical expert on the authenticity of the data in the database.

First, the core computing main system

Please refer to FIG. 4. FIG. 4 is a schematic diagram of a core computing main module according to an embodiment of the present invention.

In this embodiment, the data of the National Health Insurance Research Database (NHIRD) is used. The National Health Insurance Program in Taiwan collected medical information from more than 23 million people in Taiwan and established a National Health Insurance Research Database (NHIRD) to provide scholars with research. use. Taiwan's health insurance database covers more than 90% of Taiwan's population. In the past, Taiwan's health insurance database sampled from the parent group to establish a longitudinal health insurance database (LHID), a cancer registration file, a full hospitalization file, and a full A major injury and illness file, 1/2 child file in Taiwan, and widely used in clinical medicine, epidemiology and public health research, statistics have so far (ie in early November 2016) published more than 2,000 papers. Among the various databases provided by the Taiwan Health Insurance Database, the millions of people (LHID) are the most representative. The Million Returned Person (LHID) collects medical records for a total of fifteen years (1996-2011). The collection includes basic patient information such as outpatient visits and hospitalization records. The disease diagnosis uses the International Classification of Disease-Ninth Revision (ICD9), which includes about 928 three digit level diagnostic codes and 13813 five digit level diagnostic codes. If the E&V class code is included, that is, 1234 three-codes and 16327 five-code diagnostic codes. The data format of other databases is similar to that of millions of people (LHID). For example, the record of major illnesses in major injuries and illnesses in Taiwan is more detailed. The total number of 1/2 children in Taiwan is only included in the total period of 1996-2008 for 12 years. Half of the children in Taiwan are less than 18 or 20 years old, and the all-in-patient data file only contains information on all patients who have been hospitalized for health care for 15 years. Some famous foreign scholars published articles in famous medical journals to introduce the advantages and importance of NHIRD, that is, NHIRD is an important tool for rapid analysis of medical big data and can maintain high credibility (JAMA Intern Med.2015 Sep;175(9): 1527-9) This is the original source of basic clinical medical information for this patent.

(1) Clinical medical data collection module

The clinical medical data collection module A1 uses the central control device 10, the storage device 11 and the database 13 of the server 1 to import the database of the Taiwan Health Insurance Database (NHIRD) and other domestic and foreign medical materials, so that it can be converted into an accelerated The format of statistics, automation, and rational analysis, importing various clinical data to establish a clinical medical database. The sources of the database include millions of files (LHID), cancer registration files, full hospitalization files, and major medical illness files. The database of 1/2 children's files and other data is converted and stored. The database has obtained millions of files (LHID), cancer registration files, full hospitalization files, all major illnesses and illnesses, 1/2 children's files in Taiwan, and chronic kidney disease registration.

Referring to FIG. 4, in step A000, the system is started. In step A001, raw raw data (Raw data) obtained from various storage devices is obtained. In step A002, the clinical medical data collection module A1 processes the raw unprocessed data (Raw data) obtained from various storage devices using SAS ® (STATISTICAL ANALYSIS SYSTEM) according to the Taiwan health insurance plan. (National Health Insurance Program in Taiwan) provides the latest decoding thin decoding, cutting the original data into *.SAS7BDAT and converting to *.CSV data format, and confirming that there is no data flow during the cutting and conversion process. In step A003 and step A004, the medical record registration range of each original data, that is, the hospitalization file (DD-admission) and the clinic file (CD-OPD) are classified. In this embodiment, step A005 is to import PostgreSQL (PostgreSQL Global Development group, 9.5.2 version), SQL server 2016 (Microsoft), MariaDB version 10.1.13 using the language-structured query language (Structural Query Language-SQL). The MariaDB Corporation Ab, MariaDB Foundation) database stores and formats the cut data.

In step A006, the data stored in the database 13 is cut according to a diagnostic code different from the clinic file (CD) and the hospital file (DD). In addition to the entire hospitalization file, other databases include millions of files (LHID), cancer registration files, major injury files for all Taiwan, 1/2 children files for all Taiwan, and chronic kidney disease registration and return files. (CD) and hospitalization (DD) files, that is, only the entire hospital data file does not have an outpatient file (CD).

In step A007, the data stored in the database 13 is converted into date data. In step A007, it is first determined whether the data field is a date type data. If it is determined that the non-date type data is stored as a text type data, step A012 is performed, and if the date type data is stored as a date type data, the steps are executed. A010. In step A011, the diagnostic code (ICD) of the same field is combined with the patient identification code (ID) according to the original date. In step A006 of the diagnostic code cutting conversion, the new identification code (new ID) of the patient in the clinic file (CD) and the hospitalization file (DD) processed in step A011 is further processed, for example, performing step A008 format conversion and steps. The reverse of A009. In this embodiment, the diagnostic code input date (ie, the date when the diagnostic code is declared in the clinic or hospitalization) is merged with the diagnostic code (ICD) independent field and the format is reversed, that is, the clinic file (CD) and the hospitalization file are used. (DD) independent field clinic file (there are 5 ICD codes in the CD file, while the hospital file (DD) has three separate fields, which are processed to the same field after the format is reversed.

In this embodiment, the date type data will be used in: 1. Identifying the patient's identification code (ID_birthday, the date of birth of the patient); 2. Determining the date the patient diagnosis code is diagnosed under the clinic (func_date); The date of hospitalization (in_date) and the date of the diagnosis code (Appl_date). The field of birth date stored in the date format will be combined with the patient's identification code into a new identification code to avoid the occurrence of duplicate patient identification codes and affect the calculation results.

After the date type data is merged, it is immediately returned to step A006. After the above steps to step A009, step A013, step Á014, step A015 are executed, and the data is reordered according to the size of the date type data, and the patient's diagnosis date is pressed. Interpret the maximum value (death value) and the minimum value. The minimum value will be calculated using the date at which the patient was diagnosed at the beginning of the diagnosis and using the classification group tracked by the generation of the patient. That is, the contents of step A016 and step A017 are executed, and according to the calculation of the maximum value, the patient's last medical treatment date is interpreted, and the difference between the patient and the insured date is compared to determine whether the patient died before the medical treatment date, after the medical treatment date, or the patient Not dead.

After interpreting the patient's initial diagnosis, it follows a pre-accepted manual input (step A018), which refers to a data that can be transmitted by its website system or embedded application, in this step, especially Refers to three kinds of time lengths, which are listed as follows: 1. Exclusion period: In this period, the diagnosis of the patient is determined. An old diagnosis, that is, a disease that has been diagnosed in the past; 2. Inclusion period: When the diagnosis occurring at this time does not appear in the exclusion period, the patient is included, and if the patient meets certain conditions, You can enter the tracking period. 3. Follow up period: refers to a period in which the patient in the receiving period is tracked for a certain date and observes whether there is a real event, or whether the patient died during the period. Calculating important information such as survival rate is also the biggest difference between horizontal research and vertical type.

After accepting the date input, the patient is re-divided into outpatient (CD) and hospitalized (DD) (step A019) to facilitate the calculation of the patient's diagnosis count during the receiving period, to determine whether the patient's diagnosis is "diagnosed", to avoid inclusion The error caused by the error diagnosis (bias). The date of the visit of the patient (CD) and the hospitalized (DD) patient is compared with the date of the manual input or the length of time. The exclusion date EX is taken as an example (step A020 and step A026). If the upper limit of the set date of the exclusion date EX is less (step A026 and step A029), the patient is classified into the exclusion group; if it is greater than the set date of the exclusion date EX (step A021 step and step A027), it is simultaneously smaller than The set date of the tracking period FU, that is, the patient is classified into the receiving group (step A024 and step A030); if not, in step A022 and step A028, the patient is classified into the tracking group (step A025 and steps) A031).

When the classification of the patient is completed, the clinical medical data collection module A1 of the embodiment of the present invention enters the first loop of the nested loop (the Nested while loop), and the receiving group enters the loop of step A033, and the exclusion group Then enter the loop of step A032. The nested loop here refers to a method of longitudinally studying all the permutations and combinations in the disease matrix. For the purpose of calculation, the patient's diagnostic code is divided into the primary disease i (Index disease/primary disease/1st). Diagnosis, which will be indicated by the symbol i in the flow chart, and the disease that occurs after the primary disease is called secondary disease/2nd diagnosis, symbol j, which will be marked with the symbol j in the flowchart. Arrange the combination calculations.

(2) Automated acceleration module

After entering the nested loop, it is the scope of the automatic acceleration module A2 in the embodiment of the present invention. In the first layer of the loop in the nest loop, the starting data of the primary disease i is the smallest three or five yards in the diagnostic code (ICD), and the limit value is the largest in the diagnostic code (ICD). Three yards or five yards, when it is judged that the receiving group and the excluded group meet the loop condition, step A032 and step A033 are correct (True), that is, when the source disease i is less than or equal to the set upper limit, then the following indication is entered. Medium (step A034). In step A035, the disease code of the initial disease i of the layer is used as the initial diagnosis, and the patient in the receiving group and the exclusion group has repeated diagnosis (old diagnosis). The principle is to calculate the two using the left join culvert. Whether the group has an intersection (step A036), that is, whether the identification codes of the two groups of patients are duplicated, and the excluded patient is classified into the excluded group (Excluded) (step A037), and the patient who has entered the excluded group will not enter. Subsequent statistics. To enter the calculation of an ICD code (i++) under the loop (step A045), the flow of step A034 will be re-executed. To calculate the number of diagnoses of a certain diagnostic code of the patient who has been removed from the old diagnostics (step A038), if the number of times the patient is diagnosed in the outpatient file (CD) is greater than or equal to three times, the sentence is correctly read (True) and The patient is admitted to the experimental group (step A040), and the patient who does not meet the conditions of step A038 is interpreted as a False and the patient is admitted to the control group (step A039) . The patient who has been admitted to step A039 and step A040 merges the data of the tracking period (step A041), and enters the statistics of the tracking period data according to the setting of the baseline (step A042 and step A043). The bottom line here refers specifically to a date in the step A018 that determines the boundary between the receiving period and the tracking period.

Then enter the second layer loop (step A044), with the secondary disease j as the calculation core, the condition setting is similar to the first layer loop of the primary disease i, the flow description can refer to the above step A032 and step A033 . The second layer of variables is similar to the first layer of circling, the difference being in the second layer dealing with the death and disease occurrences.

If the error is diagnosed (True) (step A046), the number of occurrences of the disease of the secondary disease j of the layer is calculated, and the initial value of the secondary disease j is the minimum three or five yards in the diagnostic code (ICD), and the limit The value is the maximum three or five yards in the diagnostic code (ICD). When it is judged that the condition of the receiving group and the excluded group is true (True), that is, when the secondary disease j is less than or equal to the upper limit. When, the following instruction is entered (step A047). The method of step A047 is to calculate whether there is a secondary disease j in the tracking period, if there is more than or equal to one time, the reading is correct (step A049), and the event is accumulated according to the patient identification code (step A050) The date of occurrence of the secondary disease j will also be compared with the date of occurrence of the primary disease i and the person-years will be calculated (step A051). When the disease does not occur in the patient within the specified date, the next patient is counted and the patient is judged as having no event (step A048).

When the data of the person is obtained, the patient is judged to have a death event after the occurrence of the secondary disease (step A052). Refer to the three conditions of the Taiwan Health Insurance Data Base (NHIRD) for disqualification or surrender, plus other special conditions, to establish whether the automatic interpretation of patients can be classified as "non-survival" mechanism, as follows: classified into five conditions Priority is given to the three conditions of disqualification or surrender of the Taiwan Health Insurance Data Base (NHIRD); one of them is death (step A053), that is, the patient is sentenced to death by a statutory body and surrendered after three days according to the regulations of the Health Insurance Bureau. The patient, if judged as yes, is classified as dead (step A058), and if the sentence is negative, proceeds to the next level of interpretation or instruction, and so on. The second is a person who has been missing for six months (step A054), and if the reading is yes, it is classified as loss of tracking (step A059). The third is the loss of the qualification step A055, that is, the loss of the nationality of the Republic of China, the immigration of the household registration abroad, the expiration of the residence period of the foreigner, etc., if the interpretation is the same as the loss of tracking (step A060). In addition, because the Taiwan Health Insurance Data Base (NHIRD), in addition to the above three regulations of the Health Insurance Bureau, there are two situations, namely, one, military service, two, jail or detention center for more than two months. In step A056, it is judged whether the data of the Republic of China 90 years ago belongs to military service, if it is judged to be categorized as loss of tracking (step A061), and in step A057, it is judged whether the patient is in jail or detention center for more than two months. In the above, it is judged to be death (step A062). If the judgment is negative, the representative patient belongs to non-non-survival (step A063), and the surviving patient will not enter the calculation of the mortality and survival curve.

Please refer to FIG. 5. FIG. 5 is a schematic diagram of an automatic acceleration module according to an embodiment of the present invention.

In the present embodiment, the primary disease i and the secondary disease are used in the present embodiment because of the method of the longitudinal study of general clinical medical research, in which part of the method is to investigate the risk of suffering from another disease caused by a certain disease state. j represents the state of the disease, and the secondary disease j is a secondary disease caused by the disease state of the primary disease i. Entering the nested loop can achieve the purpose of calculating the combination of the primary disease i and the secondary disease, so there will be 16459(i)x16459(j) permutation combinations, which is the above-mentioned disease matrix.

Also, there is a chronological relationship between the primary disease i and the secondary disease j, that is, there is a time between the primary disease i and the secondary disease j, for example, the primary disease i → secondary disease j such as the primary disease i The disease is hypertension, and the disease of secondary disease is glycemia, then the primary disease i (hypertension) → secondary disease j (diabetes) and the primary disease i (diabetes) → j (hypertension) The meaning of the representative is not the same; if the experiment is set to the primary disease i (hypertension) → secondary disease j (diabetes), it represents the risk of developing diabetes in patients with hypertension, and the experiment is set to the primary disease i (diabetes) → secondary disease j (high blood pressure), on behalf of the study of diabetes patients with high risk of hypertension, represented by symbols, that is, injm is not equal to imjn, that is, the probability of the two is not necessarily equal, the two can not replace each other, It also means that two possible risks will exist at the same time. In addition, there is a directional relationship between the two. If the relationship between the primary disease i and the secondary disease j is calculated by the conventional method, the true disease correlation (i→j or j→i) cannot be calculated. Traditional methods such as calculating whether diabetes (i) increases the risk of hypertension (j), the calculated result is a Hazard Ratio greater than 1 and a statistically significant difference, based on which results usually result in an increased risk of hypertension due to diabetes. in conclusion. If the calculation of hypertension will increase diabetes, the calculated result is a risk ratio of about 1, combined with the above results, how to know whether diabetes actually increases the risk of high blood pressure.

In this embodiment, using the automated technique, by counting 16459 (assuming i is the primary disease, 16459 is the total number of diseases of ICD9) x16459 (assuming j is a secondary disease), the generation of the combination is tracked, and i is calculated. j (symbol: In the directionality, the link between disease i and disease j is set to L (I → J). In addition, the disease I and the disease J are exemplified below, and the disease I and the disease J are different diseases, and the correlation can be further defined according to the following description.

When the directionality of the disease I→the disease J and the disease J→the disease I is examined, the directionality of the disease I→the disease J can be known by the following formula 1.

Where i is the first disease, j is the second disease, λ is the directional judgment between the diseases, l _i→j is the relationship function value of the second disease caused by the first disease, l _j→i is caused by the second disease The relationship function value of the first disease. If λ>0, it is disease I→disease J, and if λ<0, it is disease J→disease I. In addition, these methods can be used to further understand the role of each disease in risk. In addition, the following formula 2 can be used to determine which disease is the primary disease: Λ _i = Σ _j λ _{i → j} - Equation 2

Where Λ _i is the sum of λ. When Λ _i larger, more representative of the role of disease I was the primary disease, the opposite is the second disease. The relationship between the real disease (disease I → disease J or disease J → disease I) is calculated. Similar to the concept of Social network, the value of λ is calculated from the value of λ in both directions to know the unidirectionality of the connection between nodes (nodes, vertex) and nodes, and to summarize the strength of directionality. The calculation knows the centrality of the node.

In this embodiment, a new method is provided, which is called Multiple-Nested Longitudinal Statistic (MNLS), which simplifies the steps of matrix shape longitudinal research, and its core is to use nesting. Ways to simplify the method of longitudinal research, speed up the calculation process but retain the correctness of the original statistics. This method uses the statistics of the matrix of 2x2(m*n) or above in the automation calculation, that is, the multidimensional matrix, and the operation speed of the matrix depends on the maximum value of the m element, for example, the matrix combination of 100(i)x1000(j) is used. It can effectively reduce the computing time by about one thousand times, and the matrix combination of 16459(i)x16459(j) can effectively reduce the computing time by about 16,000 times. This method is used in the calculation of big data combinations. The 1x1 disease combination or the 1x16459 combination as described above is not suitable for using the multiple nested longitudinal statistical method (MNLS), and it is necessary to achieve a 2x2 disease combination to have an acceleration effect, so the above method When the manual input (step A018) is 2x2 or more, the operation mode of the multiple nested longitudinal statistics (MNLS) is started. After the multiple nested longitudinal statistics (MNLS) is activated, it enters the nesting loop of the first layer, and whether the disease i appears in the receiving period step B001, that is, the reading is correct and the reading is positive (symbol A: index Disease positive (1st +)) Step B002, if it is positive, the patient is included in the experimental group (Exposure (+)) step B004, if the step B002 is judged as no, it represents negative (α: index disease negative (1st -)) That is, the patient is included in the control group (Exposure (-)) (step B003). Step B003 The group of step B004 respectively enters the second layer nesting loop of step B005 and step B006, and the nest loop of the layer has a judgment formula for determining whether the disease i is equal to the disease j (step B007-B008), If the disease i is equal to the disease j, then re-enter the second layer of the nest loop (step B018), and enter the calculation of the next disease j, the judgment type rescue can reduce a meaningless operation in the case of m=n. After the interpretation of step B007 and step B008 is completed, the judgment formulas of step B009 and step B010 are entered.

Step B009 is to determine whether the person suffering from the disease i has a disease j during the tracking period, that is, whether α n β m (symbol β: secondary disease positive (2nd +)), if otherwise, the patient is α nBm (symbol B) :secondary disease negative(2nd -)). Step B010 is to determine whether the person suffering from the disease i has the disease j in the tracking period, that is, whether or not An β m, if otherwise, the patient is AnBm. Therefore, in the second layer of the circle, the patient can be classified into four types, namely An β m step B012, AnBm step B014, α n β m step B011 and α nBm (step B013), and the four classification purposes are to use AnBm. The data with the α nBm category enters the statistics (step B017), while the patients of An β m and α n β m do not enter the calculation range (steps B015 and B016), and enter the second layer loop through step B018 and step B019. The relevant data of the patients classified in the second layer will be stored in the storage device 11, so that it is not necessary to reclassify the patient when entering the next disease j, and a disease on the second layer of the nest loop can be used. The patient classifies AnBm (step B014) and α nBm (step B013) until the second layer of nested loops ends and enters a disease i below the first layer of nested circles and then reclassifies the patient (step B001), and so on. . The operation speed of such a matrix depends on the maximum value of the m element, that is, the number of diseases of the second layer, and the disease matrix result counted in step B017 is as shown in step B020. Here, we create a large Trajectory database that differs from the original data, and the subsequent operations use the same Trajectory database unless the original database is replaced. That is, in the present embodiment, the plurality of data of the first clinical medical database are converted into a negative number of materials having a predetermined format, and stored in a second clinical medical database. The difference in format can be adjusted according to actual needs, and is not limited in the present invention.

In this embodiment, according to the various steps of the core computing main module, a plurality of clinical medical data and a corresponding MxN disease matrix can be generated.

Second, automated statistical module automatic regression diagnostic sub-module

Please refer to FIG. 7. FIG. 7 is a schematic diagram of an automatic regression diagnosis sub-module of an automated statistical module according to an embodiment of the present invention.

In the foregoing, it is explained that disease i and disease j calculate the risk ratio by regression analysis. In addition to comparing significant differences between different groups, regression analysis can also perform survival analysis to compare survival rates in different groups. When the survival analysis is needed, the most common mode is the cox proportional hazard model. When the Cox hypothesis is violated, it is necessary to make corrections. The Confounding factors can be used as the component variables. The commonly used method is Stratified Cox regression. The method of this patent design is flexible, allowing the user to decide which method to use, using the cox proportional hazard model or the Stratified Cox regression. In this embodiment, a more conservative Stratified Cox regression is used. In other embodiments, the former or other methods may be utilized, and the results of the regression analysis of the former or the latter method are to be verified. Therefore, the system of regression verification of the patent design automation and the method after the verification failure. The method for verifying regression analysis is called Regression Diagnostics, and the system designed in this patent is called Automatic Regression Diagnostics.

The automated system is divided into the following sections to calculate regression diagnosis: Collinearity, Independence of errors, Tests for Normality, Selective Normal Distribution Test (Optional) Tests for Normality has four parts. The first part is the diagnosis of collinearity. The result of the regression analysis of step A and step B is introduced to calculate the coefficient of determination (R2) (step D001), and it is judged whether the coefficient of determination (R2) is less than 0.8, if the correlation coefficient is too high. The representative has a collinearity problem. If the determination in step D001 is correct, the process proceeds to step D002, and so on to the method of step D004. The method of step D002 is to calculate the tolerance value (Tolerance), and the range of the numerical value is set to 0.1 to 1. The larger the value is, the lower the collinearity possibility is. If the value of the regression is within the set range, the method proceeds to step D003. The method of step D003 is to calculate the Variance inflation faction (VIF), which is actually the reciprocal of the tolerance value, so the smaller the value is, the lower the probability of collinearity is. When less than 10, the interpretation is correct and proceeds to step D004. The method. The method of step D004 calculates a Condition Index (CI), and when the condition index (CI) is less than 30, the interpretation is correct and the diagnosis of the first partial collinearity is ended. If the determination in step D001, step D002, step D003, and step D004 is an error, even if the result of the regression is judged to be collinear (step D006), the method of solving the collinearity is entered.

There are four ways to solve the collinearity. The first one is “Selecting other regression methods” (step D007), which is based on the Stratified Cox regression model of the patent. The method is a regression method. When the regression diagnosis is collinear, the regression method of the non-predetermined method is provided as the next statistical method.

The second is the use of Partial Least Squares Regression (PLS) or Principal Components Analysis (PCA). PLS establishes new Latent Variables. The prediction matrix has more variables than the observed ones. Principal Components Analysis (PCA) only explains the variation in the variables looking for influence (step D008).

The third term is Stepwise Regression and Subset Regression (Step D009), and the fourth term is to increase the number of regressions (Step D010).

The second part is to calculate the independence of the regression, using the Durbin-Watson test (DW test) (step D005) to determine whether the autocorrelation is at the level of alpha significance. Positive or negative, and test statistic d and critical values (Critical values, dL, α and dU, α), under positive values d < dU, α or under negative values (4-d >dU,α, which means that the error term autocorrelation is positive or not negative, and under positive value d>dU,α or under negative value (4-d)<dU,α, represents the error term autocorrelation is not Positive or negative. If the Dubin-Watson statistic result shows that the independence is weak, the correction is entered (step D011 to step D017). If the Dubin-Watson statistic results show strong independence, the diagnosis of independence is completed and the measurement normal distribution is entered (step D018).

The fifth item is the verification of Normal Distribution. It is detected whether the regression is abnormal or not. It is divided into necessary items and self-options. The required items are not required to be approved by the user, and are executed immediately after performing automatic regression diagnosis. It is subject to user consent or subject to user selection. The method of the mandatory item is a common statistical method, that is, the method of steps D018-D020. In step D018, the ratio of the standard deviation (SD) to the interquartile range (IQR) is calculated, and if the standard deviation (SD) = the interquartile range (IQR) / 1.35, the normal distribution is And proceeds to step D019 and step D020. If the interpretation is an error, it may be a heavier or lighter-than-normal tail, and the process proceeds to step D029. The method of step D019 and step D020 is to calculate the skewness and kurtosis. If it is a normal distribution, the skewness and kurtosis are both equal to zero.

After the process of step D001 to step D005 and step D018 to step D020 is completed and the reading is correct, that is, the result of returning to high confidence, the process may jump to step D021 to allow the user to read the returned data. If the user selects step D022 of the optional option, then the Anderson-Darling test (AD verification) (step D023), Shapiro-Wilk test (SW verification) (step D025), Kolmogorov-Smirnow test (KS verification) (Statistics of step D026) and Jarque-Bera test (step D027). When the AD verification is finished, first jump to step D024 to enter the number of people. If the number of samples is greater than 50, use KS verification step D026, and if less than 50, use SW verification (step D025) for verification, and step D025 or step D026. After execution, the Jarque-Bera test of the method of step D027 is performed. When the method of step D023 to step D027 is completed and the reading is correct, it represents the result of returning to high confidence, and the method jumps to method step D028 to let the user interpret the data of the regression. If the above determination method is non-time, the process proceeds to step D029 to automatically perform the regression diagnosis again.

Third, the verification module

Please refer to FIG. 2 and FIG. 6. FIG. 6 is a schematic diagram of a verification module according to an embodiment of the present invention.

In this embodiment, the verification module D includes an internal verification module and an external verification module, which are respectively described below.

(1) Internal verification sub-module

In order to verify that the data derived from this system is accurate and trustworthy, verification has become an indispensable and important part of the system. To this end, the verification module D of the system uses the communication device 14 of the server 1 and the data exploration technology to collect the papers that used the Taiwan Health Insurance Database (NHIRD) as a research sample, and formulates specific formats and aggregates them into a database. Compare after the day.

Since the database formed by relying on data exploration needs to be verified by a special person, and an independent group--"verification group" is specially prepared to process and verify the data of this database, the introduction of this method will explain the whole data exploration and verification. The organization of the group and the process of verification work.

The verification work can be divided into five steps. First, collect papers by means of data exploration; Second, select appropriate papers and render them into a database; 3. Check whether the database is correct by experts; 4. Collect the collected data and analyze the data.

In the first step, since the core operation result of this system is the first operation plan showing the "risk ratio" for "one disease in another disease state", in order to carry out the plan Verification, the primary screening of the paper, is to study the "Hazard ratio of IJ" as a screening condition, and set the search for the paper as a longitudinal study. In this embodiment, based on the Taiwan Health Insurance Data Base (NHIRD), although the design method is a general-purpose system, since the information is the medical record of the Taiwanese citizens, the Taiwan Health Insurance Database (NHIRD) is searched here. The paper published, that is, in step C000-1, is the selection of the Taiwan Health Insurance Database (NHIRD) as the main database, especially the paper published by the million-entry file (LHID). The academic paper search engine selects PubMed, Google Scholar, Web of Science, Medline, etc., and the journals searched by PubMed are prioritized.

In step C000-2 of the embodiment, the relevant academic papers are downloaded through the academic journals purchased by the academic institution, and the PMID of the paper (ie the paper ID used in the PubMed) is named in the portable file format (PDF). Save it. Since the papers searched by PubMed can only represent a large part of the Taiwan Health Insurance Database (NHIRD), the rest of the papers are searched by other academic paper search engines and the papers of the intersection are deleted. The portable file format (PDF) is downloaded from the downloaded Portable Document Format (PDF) via the Poppler (0.42.0 version, 2016, freedesktop.org) tool into a web page format (html) (step C001). The paper stored in the web page format (html) facilitates text exploration, and runs pandas (0.18.0 version) in python language to process web page format (html) papers (step C002). The final web page format (html) paper will be rendered in a format that is convenient for python processing (step C003). The paper of the structure will be searched for text, and the results obtained by the exploration will be stored into a database or a database. In this embodiment, the downloaded and formatted papers are stored in the database 13.

The information explored in the internal verification must have a positive impact on the verification. Therefore, in this embodiment, a small number of materials that can be used for verification in the verification are selected for exploration purposes, such as the basic data of the paper and the receiving standard. There are five major items related to exclusion criteria, grouping and stratification of patients, nature of research and statistics, and the methods of the second step are explained below. In the basic data of the first paper, in Pub C004, the PubMed ID, the title of the paper, the Digital Object Identifier (DOI), the title of the journal published in the paper, the volume of the journal (Volume, Issues) and the page have been collected. Number (Page) and so on.

The second item's acceptance criteria (step C005) and exclusion criteria (step C006), which are based on the specific requirements of the study, are as follows: Study whether diabetic patients have a higher chance of subdural hematoma than the average person (Subdural) Hematoma), therefore the "receipt criteria" (step C005) is set to patients who have found or not found diabetes during the inbox period, and the "exclusion criteria" (step C006) are patients who have had a subdural hematoma during the ingestory period. Must be excluded. There may be more than one standard in the paper, and there is not even any standard set, so the database must be processed in succession.

The third item is the grouping and stratification of patients (step C007). In the paper, patients are usually divided into several groups according to age, gender, income, social status, etc., to compare whether different groups of patients are in the disease. There are significant differences in incidence, descriptive statistics, and so on.

The fourth item is the nature of the study (step C008), namely the generational follow-up study, the case-control study, the randomized controlled trial (RCT), the meta-analysis and the systemic reviews. In this embodiment, the information that the main need to verify is the nature of the longitudinal study, so the data will be preferentially stored and processed, and the comprehensive analysis and system review may include the information of the Taiwan Health Insurance Database (NHIRD).

The fifth item is statistical data (step C009), and the data has a summary (Crude) and adjusted (Hazard Ratio), relative risk (Relative Risk), odds ratio (Odd Ratio), and corresponding Significant difference value (P value) and 95% confidence interval. The collection of statistical data mainly depends on the nature of the paper and the keywords of the data, followed by group and stratification. If the research nature of the paper is case-control research, the collected data should be the odds ratio, while the generation tracking research rules are collected. Relative risk to risk ratio. After the above information obtained by data (or text) exploration is stored in the database, it is necessary to go through the expert inspection data field, and whether the data format needs to be processed further.

The third step is for the expert to check if the database is correct. From the data (or text) exploration, we found out that the organization used the health insurance database to write the papers for the database base. There are more than 600 papers, and the titles and papers of the more than 600 papers are judged by the above conditions. And then selected several papers. In the second step, through the steps C005 to C009, the papers are collected according to a fixed database format and stored into a database (step C010). If the paper has other information written, the system will also record and store it in other items in the database. In the process of collecting, there are often some problems. For example, an academic institution cannot download the full text of the paper for free because it does not have the right to purchase the license. After the judgment, it is found that the paper does not meet the requirements of the system. The former can be downloaded through other academic institutions. The latter is initially not required to be collected, so the total number of papers collected will be less than the initial search. In addition, although the Taiwan Health Insurance Bureau is currently implementing the ICD-10 code, since it has not yet been officially launched, in this embodiment, the ICD9 diagnostic code is still collected. In the third step, the data collected in the database was over one thousand. Finally, after the expert verification, a total of hundreds of errors were found. After the amendment, a total of more than one thousand statistics were obtained. In the last step, for the collected database, the statistical data and the data calculated in this example are subjected to regression analysis (using the Hazard Ratio data as the benchmark), and the coefficient of determination (R2) is calculated. . The coefficient of determination (R2) is greater than 0.6, from which it can be known that the system developed in this embodiment is quite reliable and reliable based on the verification of such a rigorous data collection process and regression analysis. Other paper topics such as drugs, surgery, procedures, etc., which are not simple single disease-to-risk ratios for a single disease, are not included in the preliminary validation plan.

(2) External Validation Submodule (External Validation)

Since the database used in the technology described in the above method is the Taiwan Health Insurance Database (NHIRD), which is the data of Taiwan, the data of the Taiwan Health Insurance Database (NHIRD) published by Taiwan Woodland is required for verification. This is an internal verification. Due to the lack of internal verification, an external database or data is required for verification. This is external verification. If the Taiwan Health Insurance Database (NHIRD) used in the above technology is replaced by a database in the United States, the internal verification research literature is the data in the paper published by the US local database, while the external database is not the local one. Information in other US databases or in papers published in foreign countries. The externally validated local data uses the disease map developed by Taipei Medical University (The Disease Maphttp://disease-map.net/), which uses the automated statistical techniques of the medical record control method to calculate patient stratification. The data is similar to the format of this technical design. The external data from the external database uses the famous Framingham heart study (https://www.framinghamheartstudy.org/) and the HuDiNe database developed by Harvard University (http://barabasilab.neu) .edu/projects/hudine). The HuDiNe database is a similar generation tracking study in the United States where researchers use innovative methods to simulate longitudinal studies. The Felemingham Heart Study uses the generational tracking method and adjusts the interference factor according to different research topics. The experiment is designed by experts. Because the patient has been tracking the patient for several decades (beginning in 1948), there is a medical profession. Extremely high confidence, its shortcoming is limited to the heart. The stratification of the ages of the databases provided by the Fremingham Heart Study, HuDiNe, and The Disease Map varies, and is different from the default age stratification set by this technique. The age of the technology design is divided into four layers every 20 years old, and the stratification of the Disease Map is divided into ten layers every 10 years old, so to use the disease map (The The data of Disease Map) verifies the accuracy of the calculation of this technique. The layering of this technology needs to be set to be divided into ten layers every 10 years old, and the regression coefficient (R2) is calculated. The closer the coefficient of determination (R2) is to 1.0, the greater the proportion of the total variability that can be explained by this regression model, and the closer the data of the two databases are. The verification benchmark of this example is based on the decision coefficient (R2) of the Fremenming Heart Study as a priority reference standard. After the actual calculation results, the determination coefficient (R2) of the regression system of the present embodiment and the Fremingham heart research is greater than 0.6, and the system developed by the embodiment is quite reliable and trustworthy, and the international standard is very Close.

[Display device of this embodiment]

In this embodiment, the display device 12 of the cross-platform clinical medical data analysis and display system further includes a computer version of the webpage display module 121 and a mobile device version of the application display module 122. Providing the user with the cross-platform clinical medical data analysis and display system of the present invention on different vehicles can obtain the best viewing experience.

In addition, in the embodiment, the webpage content is further divided into a professional version website and a general version website, which are respectively described below. Please refer to FIG. 8 to FIG. 11 below. FIG. 8 is a flowchart of a server according to an embodiment of the present invention. FIG. 9 is a schematic diagram of a professional version webpage according to an embodiment of the present invention. FIG. 10 is a schematic diagram of a normal version webpage according to an embodiment of the present invention. FIG. 11 is a schematic diagram of an application according to an embodiment of the present invention.

(1) Professional Edition website (computer version)

In this embodiment, after the integration, statistics, and verification of the database 13, the statistical data is presented on the website. The way the site is presented, the method of data transfer (including input and output), and the presentation of the results are detailed below. The core architecture of the website is based on the Object-Relational Database Management System (PostgreSQL, PostgreSQL Global Development group, 9.5.2 version) and the Structured Query Language Server (SQL Server 2016, Microsoft) via Shiny (RStudio project.©). 2014 RStudio, Inc. for front-end interface control, the template is transmitted to the computer version of the display device 12 of the display device 12 of the server 1 and the mobile device application display module 122 for dynamic presentation (see FIG. 8 and The content shown in Figure 9).

The presentation of the results analyzed by the method of the present patent by the cross-platform of the present embodiment is based on the user's specialty, and provides different ways for the website and the data to be presented to the user. The presentation methods of the website can be divided into the following types: one for the ordinary users (or patients), the second page for the insurance practitioners, and the third for medical, pharmaceutical or public health researchers. Web page. The presentation of different web pages represents differences in display interface, usage, and presentation.

The webpage used by researchers in medicine, pharmacy or public health is the core technology of this patent. Therefore, the webpage is developed with the professional version as the core. The block F001 to block F008 shown in Figure 9 is the front interface designed by Shiny. control. Block F001 is the name of a trademark or website and appears in text or graphics. Block F002 is the data of the user account (USER ACC.). When the user account is established (step E016), the user needs to select the category to be used. The categories used are classified into the above three categories (ordinary users, insurance circles). And academics), after choosing to use the category, you need to fill in the account number, password, verification password, email (verification or result delivery), highest education, title, academic institution name (academic), company (insurance or other institution), For information such as address, if the average user adds new medical history, past surgical history, past drug history, past trauma, etc., it will be used to calculate the risk for the next 12 years. The email address will be used later for user login, billing and delivery of results. The user's profile (USER PROFILES) is stored in the Structured Query Language Server (SQL Server) and the Object-Relational Database Management System (PostgreSQL). When the user leaves the front page or re-enters after canceling the account, they need to log in again to continue. When the user logs in, the user's account (or email) and password are entered and written to the user stored in the Structured Query Language Server (SQL Server) and the Object-Related Database Management System (PostgreSQL) ( USER PROFILES) to check. User preferences, the results that have been calculated in the past will be recorded one by one. The user's preference records the user's choice of calculations, counts the number of selections and stores them in the database, and each used record is accumulated and updated in the database. After logging in, these most used records will pass the user request (USER REQUEST) and adjust the preset values of the options for subsequent processing. Users can also set it according to their own preferences. After the user logs in, the user's profile is called (step E011). When the user starts using the query system (step E006), the query parameters used by the user are transmitted to the server 1 (step E007), and according to the user. The set parameter is internally calculated (step E008). When the internal operation ends, the result is recorded in the user's side write (step E015), when the user queries the record (step E014) and accesses an item that has been calculated. As a result, the selected record will be called again (step E012). When the server receives it, it does not need to be recalculated, and the operation result is displayed on the front end interface (step E005). In block F003, the system or organ for selecting a disease, the classification of the disease is 25 combinations according to the disease classification of the diagnostic code (ICD) and the major diagnostic category (MDC). The 3 code option in the block is to dynamically change the option of 3 codes according to the choice of the system, and the 5 code option is changed according to the choice of 3 codes or the choice of the system. If the user does not select the system or organ option, The options are sorted by the diagnostic code (ICD) from small to large. The selection of block F003 has multiple selection functions, that is, researchers can select diseases such as hypertension and diabetes at the same time, and the representative can calculate the statistics under the state of "common disease" at the same time. The layout of block F004 provides the user with the choice of drug type (Drug Type, according to ATC and NHIRD code), drug dose (Drug Dose), drug dosage form (Dosage forms), drug number (Tablet), drug use site and method ( Pathway), Day Used, frequency of drug use and total dose of drug (Total Dose). The type of drug and the total dose of the drug are required to fill out the project. The calculation of the total dose of the drug is an inevitable option for the purpose of facilitating the calculation of the interval and grouping the data. According to the Taiwanese drug code search, because the Taiwan Health Insurance Bureau has established its own unique drug classification system, it is different from the internationally recognized Anatomical Therapeutic Chemical Classification System (ATC) code, NHIRD drug data. There is also a lack of Defined Daily Dose (DDD) for drugs. Therefore, if the code provided by the user is Taiwan, the following calculation method is used. The conversion of the total dose of the drug in Taiwan is equal to the single drug dose multiplied by the number of drugs multiplied by the drug frequency multiplied by the number of drug days. The calculation of ATC is in accordance with the guidelines of the WHOCC. Block E004 is also the same as block F003, with multiple selection functions, that is, researchers can select both blood pressure lowering drugs and hypoglycemic agents, which can calculate the statistics of patients' complicated medications at the same time. In addition, block F004 also provides a system for the user to calculate the relevant statistics of the surgery, based on the surgical site, surgical mode, surgical device (DEVICES) and other data for layering. The selection parameter of the front end interface of the block F003 (step E001) is transmitted through Shiny (step E005) to the layout of the block F009. Block F006 is the basic setting. This layout has a default setting when the user logs in, as shown in the figure. The layout of the basic setting block F006 is divided into the basic data grouping options, which are four items of gender, age, income, and Urbanization Level. The value of the option is the default setting of the website. Users can check different groups according to their needs. Users can also enter the interval value by themselves. The social level is a non-numeric item, so the function of inputting the interval value cannot be provided. The layout of block F006 provides a user to adjust the study period, that is, the Include time, the Exclude time, the tracking time or the Follow up time. The research category provides two methods, Cohort Study and Case Control Study. Researchers can choose different research methods according to their requirements, such as using the Taiwan Health Insurance Database (NHIRD) to study drugs and disease discovery. The most common method of risk is the medical record control study. The method of returning to the diagnosis provides a self-selecting function, that is, the method (step C018 to step C020), which is available for the user to select. Charts are produced as shown in block F006, such as Demographic, Follow Up, Kaplan-Meier Curve (KM curve), and Quantile-Quantile Plot (QQ Plot). Box Plot, Residual Analysis, Forest Plot, Cumulative incidence, Schoenfeld Plot, etc. Due to the variety of charts, they cannot be described in the figure. The above parameter after the start of the selection [E001[E002[E003] is activated by the button of the block F007 layout (step E002) and transmitted to the server (step E008). The layout of block F005 is the display of the tab control and the result page. It has the function of displaying multiple layers of pages. A total of 8 layers of front pages can be called and displayed on the front-end interface (step E005). When logging in to the system, the page displayed on the first layer is the setting page (Setting), and the interface of the block F009 is designed to display the options selected in the block F003 and the block F004, and classified according to the selected object, such as the area. As shown in block F002, the selection of the diagnostic code (ICD) will take precedence over the selection of block F004, and block F003 will be displayed above block F004 and provide a display of multiple selections. The diagnostic code display is sorted according to the size of the diagnostic code, and each diagnostic code is also accompanied by the name of the disease, and the system or organ to which it belongs, and the selected parameter (step E001) is transmitted to step E002 for standby, when the block is After F007 is started, the parameters used in step E002 will be transmitted to step E004, and the parameters of step E001 can also be transmitted through Shiny to step E005 through the front end interface for display.

Block F010 is the second level page, which is a demographic table. The demographic table is displayed according to the user-defined group, such as gender, age, etc., as well as statistics of each group, such as experiments. Exposure positive and the control negative (Exposure negative) in each group of population and percentage, the significance of each group (P value) and so on. If the user needs to count drugs or a history of comorbidities, a corresponding field will be added to the form. If the user changes the preset value when the block is checked in block F006, the system will rearrange the field according to the check box of block F006.

Block F011 is the third-level page, which is the follow-up statistics table (Follow Up). The table similar to the demographics shows the group according to the user's settings, such as gender, age, etc., and the group tracking. Statistics of events that occur, such as the Exposure positive and the Exposure negative in each group of events (Event, abbreviated as EVE), Person-year (PY), and occurrence Incinence rate (IR), etc., and Incinence Rate Ratio (IRR), adjusted hazard ratio, 95% confidence interval (95% CI) and significant difference value (95% Confidence Interval, 95% CI) P value) As above, if the user needs to count drugs or a history of common diseases, a corresponding field will be added to the form.

Block F012 is the fourth layer page, which is the Diagnostic Regression, which is the result of the display method (step C), plus Partial likelihood, Score test. And the results of statistical methods such as Wald test. If the user selects the method of the option, that is, the method of step C023 and step C025-C027, the result will be displayed on this page.

Block F013 is the fifth floor page, which is Forest Plot, which is used to display the number of people in each group, the adjusted risk ratio, 95% confidence interval and P value. The forest map will calculate based on this data. The total risk ratio, which assesses whether a disease or a drug in a population is likely to cause an increase or decrease in the risk of a disease, or an increase or decrease in mortality.

Block F014 is the sixth layer of the page, which is displayed as a picture, the picture is shown as Kaplan-Meier Curve, sub-dot, box, residual analysis, cumulative incidence and residual Analysis chart (Schoenfeld Plot) and so on. The layer layout pictures are arranged in the order of each group.

The above first to sixth layouts can be combined into a portable file format (PDF) and transmitted according to the user's wishes, based on the email address set by the user, or by allowing the user to change the email address. Go to the specified address. That is, the server 1 can provide the user with a professional version of the personal disease analysis result through the communication device 14 according to the user's needs.

Block F015 is a seventh layer page, which is set for the user to display the user's account number, password, verification password, email (verification or result transmission), highest education, title, academic institution name (academic), company (insurance) Or other institutions), address and other information, if the average user adds past medical history, past surgical history, past drug history, past trauma and other information, all the data except the account can be modified and stored by the user. The user side is stored in the Structured Query Language Server (SQL Server) and the Object-Relational Database Management System (PostgreSQL).

Block F016 is the eighth layer page, which is the About page, which is used to provide website introduction and teaching, the design and function of this patent and website, the external connection of basic statistical teaching, etc., the introduction of the laboratory and the company, etc. The content is for the user's reference.

(2) Normal Edition website (computer version)

This patent provides a way for different websites and materials to be presented to the user, including web pages used by ordinary users (or patients) (step G). The normal version of the web page inherits the features of the professional user version, but the front-end interface and data display are designed to be suitable for general users in terms of usage and data display. Step by step is used in the front-end interface design, that is, the second interface appears after the first interface is selected, and so on until all the interfaces are displayed. The front-end interface is divided into five parts, including past medical history (block G001), disease treatment (block G005), processed results (block G009), analysis of current health (block G010), and display analysis results (block G011) . After the above user creates an account, the user needs to simply enter his past medical history, where the general user system needs to record a detailed medical history for subsequent analysis. After the user logs in, the user enters the past medical history page of the block G001, and the user at the interface of the block G001 needs to select the past disease (block G002). In the choice of the disease, the system of the embodiment provides a similarity to the foregoing. The classification code (ICD) of the block F003 and the classification of the organ classification and major diagnostic categories (MDC) are selected by the general user. After the selection is completed, the diagnostic code (ICD) is not displayed as in the block F003. Here, the display system of the embodiment calls a structured query language server (SQL Server) and an object-relational database management system (PostgreSQL) in which an diagnostic code (ICD) is stored, and only obtains an diagnostic code (ICD). The name is passed back to block G002 and displayed on the front panel through Shiny. If the user is relatively healthy in the past, you can skip the project and go directly to block G005. If the user cannot recall the past disease, skip (SKIP) this item and enter block G005.

After the user selects the disease name, the display module will pop out of the window of block G003, which needs to fill in the disease about the block G002, the date of the initial symptom, and the date of the disease diagnosis. If the user cannot recall the date of the symptom or the date of diagnosis of the disease, skip the item and enter the interface of block G004.

Block G004 is to provide the user with the main symptoms of the disease or Chief Complain, Sign and other symptoms (may be written). If the user cannot recall the symptoms, symptoms, and symptoms, skip the item and enter the interface of G005.

Block G005 allows the user to fill in the treatment related to the disease filled in block G002. The field filled in the G005 interface is increased according to the number of diseases in block G002. The symptoms can also allow the user to check the relevant treatment. The selection of the G005 interface includes the drug block G006, the surgical block G007 (including outpatient or inpatient surgery) and the tracking block G008. When the user selects the drug, the window of block G006 will pop up, similar to block F004, allowing the user to select the type of drug according to the classification of ATC or Taiwan, and how to use the drug, the number of times of use, and how much to share. The total dose of the drug is displayed on the block G005 and block G006 after being converted by the system. When the user chooses surgery, the window of block G007 will jump out, according to the disease and the location of the choice, provide the name of the surgery (such as appendectomy), surgical method (traditional or Da Vinci surgical system), anesthesia (such as local anesthesia or General anesthesia), operative time (including anesthesia time), and pathology report upload. When the user selects the tracking, the window of the block G008 will pop up, and the frequency of the user's selection tracking is provided, and the tracking finds that if the user selects the option found during the tracking, the user jumps out of the window to allow the user to check the discovery. Whether it is related to the diseases listed above, if the relevant user just chooses the disease, if it is not relevant, ask the user "whether there is a need to add a new disease", if you choose to add new, jump out of the block G001, and circulate If the above steps are not added, the interface of block G009 is entered. Block G009 is the result of the user group processing. The following options are available: heal, death (user calculates other people's data), recurrence, continue treatment, continue tracking and other options.

After the ordinary user logs in again, the user's past medical record, that is, the previously filled data will be displayed in the form of a text box in the block G001, block G005, and block G009. If the user does not suffer from a new disease, the interface G001, block G005, block G009, etc. can be skipped and directly enter the interface of block G001. If the user needs to supplement the past medical history, the display module of the embodiment also provides the user to enter the interface G001 block, G005 block, block G009 and other interfaces to supplement. After the interface of the past medical history of block G001, block G005, and block G009 is confirmed, the interface of block G010 can be entered. The interface is the interface for entering the analysis before the user can analyze the future health risks based on past diseases. . There are two methods for analysis, one is to analyze the risk of a certain disease in the future, and the other is to analyze the risk of more than 16,000 diseases in the future (this number is taken as the number of diagnostic codes (ICD)), in this embodiment The latter is a charging project. In other embodiments, related results may be provided in combination with different financial services, which are not limited in the present invention.

When the user selects, the interface of block G010 will be switched to the interface of block G011, and the result of the operation will be displayed. The result displayed by block G011 is based on the selection of block G010. Since the user is a non-professional, only two types of information are displayed on the data display, such as the adjusted hazard ratio and the 10-year survival rate, and a textual explanation is provided. If the user needs to view more information, he can click the button for Expand for more data, and the expanded data includes Case fatality (the ratio of deaths among patients suffering from a disease). Conditional Probability, Cumulative Survival Probability, etc. at a certain point in time. If the user chooses the risk of a disease, it only shows the risk of a disease and the 10-year survival rate, that is, the patient chooses to analyze only the risk of hypertension in the next 10 years, only the risk of hypertension and 10 years of survival. rate. If the user chooses to analyze the risk of more than 16,000 diseases in the future, it will display more than 16,000 diseases and can be reordered according to the risk or survival rate. If the user needs advice after viewing the results, they can check the suggestion system of this website. The proposed system of this website, based on the system of the disease and the address of the patient's department, finds the nearest hospital to provide user enquiries and assists the user in contacting the department's physicians and appointments. The result of block G011 can also be printed. Export or send to an email (for example: user's email or family physician's email) for use with a regular version of the individual's disease analysis results.

(3) Mobile device version application

The cross-platform clinical medical data analysis and display system of the embodiment of the present invention can also provide the system webpage content or application suitable for display on the mobile device through the mobile device version application display module 122 of the display device 12 of the server 1. The following is described according to the application.

In order to be applied to smart phones and medical instruments, the mobile device version application display module 122 in this embodiment uses the object-oriented high-level programming language C# (6.0 version, Microsoft) based on the .NET framework to create iOS, Android, and Windows. Mobile application (app) for mobile devices such as Phone. The app can be downloaded from a digital media web store (such as the iOS itune store) or built into the phone, and can be used without downloading. In this embodiment, the design of the application system refers to the previously described website system, inherits two major systems including ordinary users and professional users, and the interface simplifies the layout of the front-end interface of the webpage, and buttons and tick buttons. The text box and font size are increased, which is convenient for the user of the mobile device to watch or select with a finger. In this embodiment, all interfaces of the application have a scroll function that slides vertically, so that the user does not have to use the data or consult the data because of insufficient space. All interfaces have the ability to query instructions. The conversion between the application interfaces is a step by step method. Press the next page button to go to the next interface. In addition, the application is divided into two types: the ordinary (patient) version of the application and the academic (professional) version of the application. The following describes the design and function of the interface. The above two applications inherit the user account and the login and logout interface of the block F002. After the user account is established (step E016), the data is uploaded to the server 1 and stored in the structured query language server. (SQL Server) and object-relational database management system (PostgreSQL), check when the user re-login. After the user logs in, the user's profile will be called (step E001). When the user starts using the query system (step E006), all selected parameters are transmitted to the server 1 (step E007), and according to the parameters set by the user. Perform an internal operation (step E008). After the calculation is completed, the server 1 provides the user with an analysis result through the communication device 14.

(4) Mobile device version application (normal version)

The interface of the normal application is the same as the normal version of the website system. It is divided into five parts, including the past medical history (block H001, which is the G001 interface of the web version), and disease processing (block H002, the web version). Block G005 interface), processed result (block H003, ie the G009 interface of the web version), analysis of the current health (block H004, ie the G010 interface of the web version), display analysis results (block H005 to district) Block H006, the block G011 interface of the web version). Block H001 to block H003 interface retains the function of block G002 to block G004 and block G006 to block G008 of webpage, and after the data registration is completed, it enters the interface of block H004, providing user selection. Calculation service. The result of the internal operation is transferred to the user's mobile phone for storage and also recorded in the user's profile (step E015). When the user queries the record (step E014) and accesses the result of an operation. The selected record is called again (step E012). When the server 1 receives it, it is judged that the recalculation is not required, and the operation result is displayed on the front end interface (step E005). The result of the storage will be displayed on the front-end interface when the interface of block H005 or block H006 is clicked. The interface of block H004 has two computing functions, which provide the user with the risk of calculating any disease in one case, and the other function is to calculate the risk of all the diseases of more than 16,000 diseases; The data filled in by the block H001 and the block H003 interface and the disease range selected by the block H004 will be transmitted to the server 1 (step E007) to perform the operation (step E008), and the result is output to the area. The interface between block H005 and block H006. The interface of block H005 shows the risk of a certain disease, and the interface of block H006 shows the risk of all diseases. The display of the data has a scroll function that slides vertically.

(5) Mobile Device Edition (Professional Edition)

The interface of the professional application is the same as the professional version of the website system. The layout is designed to be simplified into 4 parts, namely the tab control (block H008) and the trademark display layout (block). H009), workspace layout (block H010), text description area (block H011).

The layout of block H008 is the control of the Tab Control control setting and the result page. It has the function of displaying multiple pages. In order to simplify the interface and reduce the use space, the block H005 layout inherits the block F005 layout and merges the blocks at the same time. F003-F004 and the layout and function of block F006-F007. The layout of block H008 has 12 layers of front-end pages that can be called and displayed in the block H010 layout of the front-end interface. The 12-layer layout is: the first layer uses teaching (block H012), and the second layer inputs diagnostic code ( ICD) (block H013), input drug (block H014), basic setting and operation mode (block H015), behavior (block H016), population ratio result (block H017), tracking and observation (block) H018), regression diagnosis result (block H019), forest map (block H020), other map (block H021), user setting (block H022), and (block H023).

The first layer uses the teaching interface (block H012), detailing how the system is used, allowing the user to skip this step. The second layer inputs the diagnostic code (ICD) layout (block H013), the third layer enters the drug details layout (block H014) and the fourth layer basic setting and operation mode layout (block H015), respectively. The function of block F003, block F004 and block F006. The layout block H016 inherits only the execution function of the layout block F007. The parameters of each layout (step E001 to step E003) of each layout of the block H012 to the block H015 are immediately stored in the storage device 11 until the execution function of the block H016 is started, by step E002 The operation is transmitted to the server 1 (step E008), and the result of the calculation is transferred to the memory of the mobile device. The data is displayed in the interface from block H017 to block H021.

Population ratio results (block H017), tracking and observation results (block H018), regression diagnosis results (block H019), forest map block H020 and other maps (block H021), respectively, inherit the area of the professional version of the web page The layout of block F010, block F011, block F012, block F013, block F014 and its functions. The last two layers of the tab control (block H022 to block H023) inherit the design and function of the block F015 and block F016 of the professional version of the webpage respectively. .

[Possible effects of the examples]

In summary, the cross-platform clinical medical data analysis and display system of the present invention performs further analysis and statistics through the data of a clinical medical database, which greatly reduces the calculation amount, speeds up the processing, and can not reduce the accuracy thereof. In addition, the professional or general version of the analysis results can be obtained from the webpage or application provided by the system, which not only provides users with clear and rapid medical analysis information, but also instantly understands the possible development direction of their health. .

The above description is only an embodiment of the present invention, and is not intended to limit the scope of the invention.

Claims (10)

 A cross-platform clinical medical data analysis and display system is provided in a server, wherein the server comprises a central control device, a storage device and a communication device, and the cross-platform medical data analysis and display system comprises: a core computing master The module collects and generates a plurality of clinical medical data, including: a clinical medical data collection module, and the disease diagnosis code, the judgment of various dates, and the medical treatment according to the plurality of medical materials of the at least one first clinical medical database Sorting and collecting data, and converting a plurality of medical materials of the first clinical medical database into a plurality of predetermined format data, wherein the clinical medical data collecting module stores the plurality of predetermined format data in a second clinical medical a database; and an automated acceleration module for performing a disease relationship analysis based on a multiple nested longitudinal statistical method and the plurality of predetermined format data in the second clinical medical database, and generating a disease matrix, wherein The disease matrix is an MxN matrix; an automated statistical module, utilized to A regression diagnosis method for statistically analyzing data of the core operation main module to generate an analysis result; and a verification module, at least according to the plurality of papers corresponding to the clinical data of the first clinical medical database The data is verified by the disease matrix of the core operation main module and the analysis result.    The cross-platform clinical medical data analysis and display system of claim 1, wherein the verification module comprises an internal verification sub-module and an external verification sub-module, wherein the internal verification sub-module is corresponding to a plurality of thesis data of the clinical data of a clinical medical database, verifying the operation result of the core operation main module and the analysis result provided by the automatic statistical module, wherein the external verification sub-module The plurality of data materials are verified according to the operation result of the core operation main module of the second clinical medical database and the analysis result provided by the automated statistical module.    For example, in the cross-platform clinical medical data analysis and display system of claim 1, wherein, in the disease matrix, M is greater than or equal to 2, and N is greater than or equal to 2.   The cross-platform clinical medical data analysis and display system of claim 1, wherein the multiple nested longitudinal statistical method determines the relationship between different diseases in the disease matrix according to at least one of the following formulas, wherein The disease matrix includes a primary disease as well as a single disease: Λ _i =Σ _j λ _{i → j} - Equation 2; wherein i is the first disease, j is the second disease, λ is the directional judgment between the diseases, and l _i→j is the first disease causing the second disease The relationship function value, l _j→i is the relationship value of the first disease caused by the second disease, Λ _i is the sum of λ, and the calculation Λ can know the central characteristic of the node.  For example, the cross-platform clinical medical data analysis and display system of claim 4, wherein when λ is greater than 0, the second disease is caused by the first disease, and when λ is less than 0, the first disease is caused by the second disease.    The cross-platform clinical medical data analysis and display system of claim 1, wherein the server further comprises a display device, the display device comprising a computer version of the webpage display module and a mobile device version of the application display module .    The cross-platform clinical medical data analysis and display system of the cross-platform clinical medical data analysis and display system, wherein the cross-platform clinical medical data analysis and display system displays a computer version webpage through the computer version of the webpage display module and the communication device The cross-platform clinical medical data analysis and display system provides the information required by an application in a mobile device of the user through the mobile device application display module and the communication device.    For example, in the cross-platform clinical medical data analysis and display system of claim 7, wherein a user obtains a general version of the personal disease analysis result or a professional version of the personal disease analysis result through the computer version of the webpage.    For example, in the cross-platform clinical medical data analysis and display system of claim 7, wherein a user obtains a general version of the personal disease analysis result or a professional version of the personal disease analysis result through the application.    For example, the cross-platform clinical medical data analysis and display system of claim 1 includes the Cox proportional risk model or the layered Cox regression model.