Method for analyzing patient information, patient information analysis device, patient information analysis program, and recording medium.

Abstract

To provide a technical method that makes it possible to search for research perspectives from visualized data and to easily select and analyze appropriate data to be analyzed.SOLUTION: A patient information analysis method includes the steps of: displaying in a diagram the order of treatments on a first axis, a type of treatment method on a starting point side of the order of treatments on the first axis and a result or halfway state of treatments on an ending point side of a second axis on the basis of electronic medical record data of multiple patients; selecting one or more patient groups consisting of the multiple patients with a common treatment in the diagram display; and analyzing the electronic medical record data related to the patient group when the patient group is selected.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a patient information analysis method, a patient information analysis apparatus, a patient information analysis program, and a recording medium. Background Art

[0002] Along with the development of digital technology in recent years, various types of Real World Data (data such as electronic medical records obtained at clinical sites, hereinafter referred to as "RWD") have been accumulated, and expectations for the utilization thereof are increasing. In particular, in the field of pharmaceutical development, the utilization of RWD is being explored, and cases where RWD has been used as approval application data for rare diseases overseas have also been reported. In the field of pharmaceutical and therapy development, RWD refers to data generated and collected in daily medical practice, such as electronic medical record data and medical expense data. In a broader sense, RWD can also include data collected from self-monitoring devices and the like, as well as patient registry data. Electronic medical record data has the characteristic of enabling detailed tracking of outcomes of past patient treatments. As an example of utilizing electronic medical record data, analyzing a patient's patient journey can be expected to contribute to supporting medical care for healthcare workers and supporting the research and development of drugs and medical devices.

[0003] Here, a patient journey refers to a representation of the actions, thoughts, emotions, etc., of a patient in the process of progressing through treatments such as medical consultations and medication after the patient recognizes a disease or symptom. It encompasses the period from acute-phase medical care to the recovery phase and life in the community during the chronic phase, and is used as a term to describe the process of living as a "person living with an illness" in a medical institution and the local community. The patient journey is gaining consensus among pharmaceutical companies and healthcare institutions as a way to provide medically, psychologically, and behaviorally sound healthcare services based on patients' healthcare-seeking behaviors. In other words, by visualizing the process patients go through from the pre-symptomatic stage to treatment, it becomes possible to understand, for example, when and why optimal treatment is not being reached. Specifically, it becomes possible to consider what advice and policies doctors and medical staff should propose at turning points such as the time of consultation, the start of treatment, and discharge, and by analyzing the intentions behind them in detail and adding quantitative data, companies can identify opportunities for intervention while contributing to the healthcare system.

[0004] As a method and technology for utilizing such patient journeys, Patent Document 1 discloses a technology that provides a treatment route analysis and management platform by showing the results for treatment routes for similar patients as a Sankey diagram. Furthermore, Patent Document 2 discloses a technology that inputs medical data such as vital sign information, performs pattern classification, displays the results in a Sankey diagram, and when a plot is selected by the user, displays the results of the classification process used to calculate the performance indicator corresponding to that plot. Furthermore, Non-Patent Document 1 provides interactive web tools for the simple analysis of clinical and survival characteristics of esophageal cancer patients, including graphs such as bar plots, Sankey plots, line plots, and maps. Furthermore, Non-Patent Document 2 describes the use of heatmaps and Sankey diagrams as tools for visualizing longitudinal clinical, cancer genomics, and molecular biology data, supporting interactive exploration and ranking of clinical and molecular characteristics. [Prior art documents] [Patent Documents]

[0005] [Patent Document 1] Japanese Patent Publication No. 2020-042761 [Patent Document 2] Japanese Patent Publication No. 2021-039491 [Non-patent literature]

[0006] [Non-Patent Document 1] ECCDIA: an interactive web tool for the comprehensive analysis of clinical and survival data of esophageal cancer patients Jingcheng Yang, Jun Shang, Qian Song, Zuyi Yang, Jianing Chen, Ying Yu & Leming Shi [Non-Patent Document 2] OncoThreads: visualization of large-scale longitudinal cancer molecular data Theresa A Harbig, Sabrina Nusrat, Tali Mazor, Qianwen Wang, Alexander Thomson, Hans Bitter, Ethan Cerami, Nils Gehlenborg [Overview of the project] [Problems that the invention aims to solve]

[0007] However, the technologies disclosed in Patent Documents 1 and 2 both analyze the patient journey based on negative data (adverse events, hospitalizations, etc.) and do not display or analyze information from the entire medical record. Furthermore, the tools provided in Non-Patent Documents 1 and 2 do not handle detailed medical data such as electronic medical record data, nor do they integrate visualization and analysis functions.

[0008] In other words, in conventional technologies, including the aforementioned literature, there was a problem in that users could not explore research perspectives from visualized data and select and analyze appropriate data for analysis. This invention has been made in view of the above-mentioned problems, and aims to provide a technical method that enables the exploration of research perspectives from visualized data and facilitates the selection and analysis of appropriate data for analysis. [Means for solving the problem]

[0009] The present invention provides a method implemented on a computer for analyzing patient information, comprising: displaying a diagram on a first axis showing the sequence of treatments based on the electronic medical record data of multiple patients; on a second axis showing the type of initial treatment on the starting side of the sequence of treatments on the first axis, and the final treatment status or intermediate treatment status on the ending side; allowing selection of one or more patient groups consisting of multiple patients with common treatments in the diagram display; and analyzing the electronic medical record data relating to the selected patient group.

[0010] More specifically, this disclosure encompasses the following aspects: [Aspect 1] A method implemented on a computer for analyzing patient information, Based on electronic medical record data from multiple patients, the first axis displays the sequence of treatments, and the second axis displays the treatment method on the starting point side of the sequence of treatments on the first axis, and the treatment result or intermediate state on the ending point side, in a diagram. In the diagram display, it is possible to select one or more patient groups consisting of multiple patients who share the same treatment, When the aforementioned patient group is selected, the electronic medical record data relating to that patient group is analyzed. A method for analyzing patient information, including the patient's information. [Aspect 2] The patient information analysis method according to aspect 1, further comprising creating a machine learning model that predicts the classification of the two selected patient groups using the electronic medical record data of the selected patient groups when two patient groups consisting of multiple patients with common treatments are selected in the diagram display. [Aspect 3] The patient information analysis method according to aspect 2, further comprising identifying factors that contribute to the classification of patient groups predicted by the machine learning model created. [Aspect 4] The patient information analysis method according to aspect 1, further comprising determining a significant difference between the two selected patient groups for at least one data item of the electronic medical record data of the selected patient groups when two patient groups consisting of multiple patients with common treatment are selected in the diagram display. [Aspect 5] The method for analyzing patient information according to any one of aspects 1 to 4, wherein the patient is a patient with ovarian cancer. [Aspect 6] A method for analyzing patient information according to any one of aspects 1 to 5, wherein the treatment includes at least one selected from the group consisting of drug therapy, surgery, radiotherapy, palliative care (best supportive care), and observation. [Aspect 7] A method for analyzing patient information according to any one of aspects 2 to 4, wherein two patient groups are selected: a group of patients with a good prognosis and a group of patients with a poor prognosis. [Aspect 8] The method for analyzing patient information according to aspect 3 for identifying factors that affect the treatment completion rate. [Aspect 9] A patient information analysis device for analyzing patient information, A diagram display means that, based on the electronic medical record data of multiple patients, displays the sequence of treatments on the first axis, and on the second axis, the type of treatment on the starting side of the sequence of treatments on the first axis, and the result or intermediate state of treatment on the ending side, In the diagram display shown by the diagram display means, a patient selection means is provided to allow the selection of one or more patient groups consisting of multiple patients who share the same treatment. A patient information analysis device characterized by having a patient information analysis means that analyzes electronic medical record data relating to a patient group when the patient group is selected by the patient selection means. [Aspect 10] The patient information analysis apparatus according to Aspect 9, further comprising machine learning model creation means for creating a machine learning model that predicts classification of the two selected patient groups using electronic medical record data of the selected patient groups when two of patient groups each consisting of a plurality of patients sharing a common treatment are selected in the diagram display displayed by said diagram display means. [Aspect 11] The patient information analysis apparatus according to Aspect 10, further comprising contributing factor analysis means for analyzing factors contributing to the separation of the two selected patient groups using the machine learning model created by said machine learning model creation means. [Aspect 12] The patient information analysis apparatus according to Aspect 9, further comprising significant difference determination means for determining a significant difference between the two selected patient groups for at least one data item of the electronic medical record data of the selected patient groups when two of patient groups each consisting of a plurality of patients sharing a common treatment are selected in the diagram display displayed by said diagram display means. [Aspect 13] A program that causes a computer to function as each means in the patient information analysis apparatus according to any one of Aspects 9 to 12. [Aspect 14] A computer-readable recording medium having recorded thereon a program for causing a computer to function as each means in the patient information analysis apparatus according to any one of Aspects 9 to 12. Effects of the Invention

[0011] According to the present invention, a user can easily perform selection and analysis of appropriate analysis target data from a research perspective explored from visualized data by linking a visualization function and an analysis function. More specifically, a patient journey of treatment transition can be visualized, and exploratory comparative verification and statistical analysis can be performed on patient groups. Brief Description of the Drawings

[0012] [Figure 1] It is an example of a diagram display according to the method of the present invention. [Figure 2] This figure is a diagram showing an example of a state where two treatment paths are designated by a user and displayed distinguishably from each other. [Figure 3] This figure is a diagram showing an example of a screen that displays results of analysis of a selected patient group. [Figure 4] This figure is a diagram showing a patient narrowing-down process. [Figure 5] This figure is a functional block diagram showing a schematic functional configuration of a patient information analyzer according to an embodiment of the present invention. [Figure 6] This figure is a flowchart showing a flow of judgment and processing in the patient information analyzer according to an embodiment of the present invention. [Figure 7] This figure is a diagram showing a schematic configuration of a computer that can implement the patient information analyzer according to an embodiment of the present invention. [Figure 8] This table shows factor analysis results. MODE FOR CARRYING OUT THE INVENTION

[0013] Hereinafter, a patient information analysis method according to an embodiment of the present invention will be described with appropriate reference to the drawings. (First Embodiment) The patient information analysis method according to the first embodiment of the present invention is based on electronic medical record data of a plurality of patients, and displays a diagram in which: the order of treatments is set on a first axis; on a second axis, different treatment methods are arranged on a starting point side of the treatment order on the first axis, and treatment results or intermediate states are arranged on an end point side. In the diagram display, one or more patient groups each consisting of a plurality of patients sharing a common treatment can be selected, and when the patient group is selected, electronic medical record data related to the patient group is analyzed.

[0014] An example of the diagram display is shown in FIG. 1, but the present invention is not limited thereto, and diagram display methods that can express the order of process flow and the flow of time, such as a parallel coordinate plot, a polar area diagram, and a stacked bar chart, may be employed. The diagram shown in Figure 1 is a so-called Sankey diagram, which can represent the sequence and flow rate of a process. Figure 1 is an example that focuses on drug therapy information for ovarian cancer, with the horizontal axis 101 representing the number of cycles 103 representing the sequence or number of drug therapies for that disease (ovarian cancer), and the vertical axis 102 representing the different drug therapy regimen names 104. It should be noted that the use of a Sankey diagram is merely an example, and any display format such as a flowchart may be adopted depending on the situation.

[0015] Furthermore, node 105 represents each cycle of therapy, and links 106 that cross the nodes show the changes in the patient's clinical data over time ("treatment history" or "treatment path"). The change in the thickness of each link represents the change in the number of patients. Please note that the disease "ovarian cancer," treatment method "drug therapy," and data transition "treatment history" are merely examples. Other cancers such as "colorectal cancer," "lung cancer," and "stomach cancer," other diseases such as "diabetes," "hypertension," and "myocardial infarction," treatment methods such as "surgery," "radiation therapy," "immunotherapy," "dietary therapy," and "palliative care," and data transitions such as "clinical test values," "tumor size," "cancer staging," "performance status," and "vital data" can also be included.

[0016] In the method of the present invention, when one or more links are indicated by a user through pointer operations such as mouse clicks, a group of patients who share a common treatment history corresponding to those links is selected from the data. The links corresponding to that patient group (the pointed-to links) are then displayed differently from other links, for example by using a different color, so that they are particularly distinguishable from other parts in the diagram display.

[0017] Figure 2 shows an example of a situation where two treatment paths (corresponding links) are indicated by the user and are clearly distinguishable from each other. In other words, in the example in Figure 2, the group of patients with a good outcome is designated as Group A, and the group with a poor outcome is designated as Group B, and each is clearly distinguishable from the others.

[0018] Figure 3 shows an example of a screen display that appears after a user has performed an action, such as selecting a menu item, while a link has been selected in Figure 2. In the example in Figure 3, Figure 301 displays statistical values ​​such as the maximum, minimum, and mean values ​​for each item in the clinical data immediately preceding Cycle 1 for the selected patient group. Histogram 302 displays a histogram of the selected clinical data when cell 1 in Figure 301 is selected by a user's mouse click or other point operation. In the example in the figure, the number of patients for each test item in the patient group is displayed.

[0019] Thus, in the first embodiment of the present invention, the user can easily select appropriate data to be analyzed from visualized data (Sankey diagram in this example) and from explored perspectives (goodness or badness of the progress in this example), and perform analysis of the selected data (number of patients for each test item in this example).

[0020] (Second embodiment) A second embodiment of the present invention provides a method for analyzing patient information in which, when two patient groups consisting of multiple patients with common treatments are selected in the diagram display, a machine learning model is created that predicts the classification of the two selected patient groups using the electronic medical record data of the selected patient groups.

[0021] Electronic medical records manage various clinical data, including blood test data such as white blood cell counts and albumin levels, daily vital data such as heart rate, and profile data such as height and weight. This method uses this clinical data to create a machine learning model that predicts the classification of two patient groups. Here, "machine learning" refers to an algorithm that automatically improves through learning from experience. It learns using data called "training data" or "learning data," and uses the learning results to perform classifications such as rules. For example, using a machine learning method called logistic regression analysis, several clinical data can be classified based on the probability of a certain outcome occurring. This method makes it possible to classify patient groups using electronic medical record data. (Third embodiment) The patient information analysis method according to the third embodiment of the present invention identifies factors that contribute to the classification of patient groups predicted by the machine learning model created. This method explores the factors that differentiate two patient groups, which are classified using a machine learning model created with clinical data managed in electronic medical records, such as blood test data including white blood cell counts and albumin, daily vital data including heart rate, and profile data including height and weight. Specifically, factor analysis can be performed using logistic regression, a binary classification model in machine learning, but it is not limited to this. Classification using algorithms such as decision trees, neural networks, and ensemble learning, neural networks that do not require training data, clustering such as k-means clustering, and other machine learning methods may also be employed.

[0022] Furthermore, a "binary classification model" is a classification model that takes only two values, such as 1 or 0, or positive or negative, to indicate whether the treatment outcome is good or bad. Furthermore, "logistic regression" is a statistical method for predicting the probability of a certain event occurring (for example, the rate of product failure or the probability of improvement in a medical condition).

[0023] "Prediction" clarifies the cause-and-effect relationship between certain causes and the resulting outcomes. The data on the cause side is called the "explanatory variable" or "independent variable," and the variable on the outcome side is called the "dependent variable" or "response variable." In this method, the dependent variable is two patient groups selected by the user, and electronic medical record data is used as candidate explanatory variables. Classification models are created using various combinations of explanatory variables, and the best model is searched for. The combination of explanatory variables entered into the best model is then identified as the influencing factor.

[0024] The variable augmentation method is employed as the search method. That is, the number of variables is increased one by one from zero, and the variable that best represents the model at each step is selected. The model is evaluated using the Akaike Information Criterion (AIC) and the prediction accuracy of the classification model obtained through cross-validation (K-hold Cross-Validation). Here, AIC is an index that evaluates the degree to which the model fits, and a smaller evaluation value indicates a better model. If the AIC evaluation value is higher than that of the previous step, it is determined that no better variables can be found, and the search is terminated. Furthermore, cross-validation is performed on the combination of variables selected using AIC at each step, and the selected variables are evaluated by the model's prediction accuracy (Accuracy). AIC evaluates using a single model, while cross-validation (K-hold Cross-Validation) evaluates the generalization performance using K models. If the model's prediction accuracy deteriorates compared to the previous step, it is determined that overfitting has occurred, and the search is terminated.

[0025] Variables that are correlated with variables selected using AIC tend to be less likely to be selected in later steps. Therefore, to prevent highly correlated variables from being excluded due to the nature of AIC evaluation, we also add the top three variables with the highest correlation to the selected variables as factors.

[0026] The effectiveness of this method was verified using clinical data of ovarian cancer patients at the National Cancer Center Hospital East from May 1, 2013 to November 1, 2020, which was approved by the Ethics Review Committee of the National Cancer Center. The data was anonymized to prevent the identification of individual patients. However, the above data included records of patients who were diagnosed with suspected ovarian cancer, as well as data of patients who did not receive treatment (and therefore have no treatment records) because they visited the clinic for a second opinion. Therefore, we narrowed down the patient list according to the process shown in Figure 4.

[0027] Specifically, in case 401, to narrow down the search to patients with ovarian cancer, the study was limited to patients whose disease names were stored under ICD codes C56, C57, C570, C578, C579, and C482. Here, ICD codes are International Classification of Disease codes recommended by the WHO (World Health Organization) for international comparison of injury and death statistics. In case 402, to exclude patients suspected of having ovarian cancer, the system was limited to patients whose disease diagnosis flag was set to "confirmed." In case 403, clinical trial patients who were not receiving regular medical care were excluded. In case 404, to visualize treatment history, we limited the analysis to patients with treatment records. Specifically, we extracted patients whose path names, recorded in surgical records or regimens (chronological treatment plans specifying drug dosages, methods of use, and treatment duration, or types of drug therapy), contained the words "ovary" or "peritoneum."

[0028] These four steps narrowed down the number of patients to be used in this verification to 112. Then, we visualized the ovarian cancer treatment history of the patients narrowed down as described above.

[0029] There are several drug therapy regimens for ovarian cancer, including dose-dense carboplatin + paclitaxel (hereinafter referred to as "dd-TC therapy"), which uses paclitaxel and carboplatin, and it was also observed that patients switched to a different regimen during the course of treatment. One cycle was defined as a combination of administration days and drug-free days for each regimen, and the patient journey of treatment progression was stratified for each cycle using a diagram according to the method of the present invention.

[0030] In this study, dd-Weekly-TC therapy for ovarian cancer is considered standard treatment with up to 6 cycles of continuous treatment, and this group of patients is presumed to have a favorable outcome. On the other hand, patients who discontinued treatment after 5 cycles or less are presumed to be a group of patients with poor outcomes, either due to progression of ovarian cancer during treatment or discontinuation of treatment due to side effects, etc. By statistically comparing and analyzing the clinical data of these favorable / unfavorable outcome groups, it is thought that some clinical factors contributing to either a favorable or unfavorable outcome can be identified. Therefore, we selected two patient groups for dd-TC therapy: a group of patients with a good outcome (those who completed the treatment with a good outcome) and a group of patients with a poor outcome (those who completed the treatment with a poor outcome). These groups were then used as data for factor analysis using a machine learning classification model.

[0031] The target population for this study consisted of 49 patients who had received dd-TC therapy at least once. Within this group, a favorable outcome was defined as completing 6 consecutive cycles of dd-TC therapy without receiving any other treatment. A poor outcome was defined as discontinuing dd-TC therapy after 5 cycles or less, or completing 7 or more consecutive cycles. Based on these definitions, the dependent variable was determined to be 25 patients in the favorable outcome group and 24 patients in the poor outcome group. Prior treatment history before the application of dd-TC therapy was not considered in determining the dependent variable. The candidate explanatory variables were 341 types of data—laboratory values, vital signs, and profile data—that were recorded at least once in the electronic medical records of the patient group (the dependent variable). The data used were those immediately prior to the start of dd-TC treatment, and the problem was designed to predict the post-treatment course based on the pre-treatment laboratory values, vital signs, and profile.

[0032] First, among the laboratory values ​​obtained by patients treated with dd-TC, there were 0% missing values, and all 49 patients in the analysis underwent testing for 39 variables, which were used as candidate explanatory variables for factor analysis. As a result, serum potassium (K) and serum urea nitrogen (UN) variables were identified as having a high correlation with favorable / unfavorable outcomes. Figure 8 shows the variables (K, UN) selected through factor analysis, the variables with high correlation to those variables, and the correlation coefficients.

[0033] Next, a data-driven factor analysis was conducted using 65 types of test values ​​with less than 25% missing data. The top two variables were K and UN, which were consistent with the variables tested by all 49 participants in the population. In other words, it was estimated that K and UN, which are related to indicators of renal failure and adrenal insufficiency immediately before the start of treatment, are factors that influence the success or failure of the treatment.

[0034] Thus, according to the method of the present invention, users such as doctors can visually grasp how many patients switched to which treatment method at which stage, and can easily investigate what kind of patient groups with what treatment histories can be explored from what perspective. Furthermore, the distribution and statistical differences between the two patient groups of variables identified through factor analysis can be easily confirmed. Furthermore, by integrating interactive features for selecting patient groups to analyze with statistical analysis functions such as calculating statistics and displaying histograms, the complexity of data analysis is eliminated. Furthermore, from the perspective of both medical professionals and patients, visualizing the patient journey based on medical record data would allow patients (including those suspected of having cancer) to easily understand what kind of treatment they will receive and what course their condition will follow, enabling them to choose treatment that better reflects their own wishes.

[0035] (Fourth embodiment) The patient information analysis method according to the fourth embodiment of the present invention determines, in the diagram display, a significant difference between the two selected patient groups for at least one data item of the electronic medical record data of the selected patient groups when two patient groups consisting of multiple patients with common treatments are selected. Here, "significant" means that the difference between the "hypothesis" and the "actually observed result" is not merely an error. In other words, if a data item (for example, K: serum potassium) is determined to have a significant difference between, for example, a group of patients with a good prognosis and a group of patients with a poor prognosis, then it can be inferred that this data item is a factor influencing the success or failure of treatment.

[0036] According to the method of the present invention, users such as physicians can interactively select two patient groups to be analyzed and one or more data items, and easily confirm the statistical differences between the two patient groups for the selected data items.

[0037] (Fifth embodiment) The method for analyzing patient information according to the fifth embodiment of the present invention is particularly applicable to patients with ovarian cancer. Ovarian cancer is a malignant tumor that develops in the ovaries. In addition to cancer, which is a "malignant tumor," tumors that develop in the ovaries (ovarian tumors) include "benign tumors" and "borderline malignant tumors" which have properties intermediate between malignant and benign.

[0038] Treatment options are determined based on the stage of cancer progression, the nature of the cancer, and the patient's physical condition. Because the ovaries are located deep within the pelvis, it is difficult to accurately assess the extent of cancer spread without examining the ovaries removed during surgery. Therefore, the stage of ovarian cancer is generally determined after surgery. However, surgery is often hesitated because removing the ovaries makes it difficult to preserve fertility.

[0039] Therefore, by visualizing the patient journey, if we can statistically estimate the future course of the disease from medical record data before or during treatment, we can select the optimal treatment, which would be a great boon, especially for patients who hope to have children in the future (including those suspected of having cancer). In other words, the method of the present invention has particularly advantageous effects on patients with ovarian cancer.

[0040] (Sixth embodiment) A method for analyzing patient information according to a sixth embodiment of the present invention is characterized in that the treatment includes at least one selected from the group consisting of drug therapy, surgery, radiotherapy, palliative care, and observation. For example, cancer treatment includes drug therapy (chemotherapy), surgery, and radiation therapy, and each treatment method is used individually or in combination, taking advantage of its unique characteristics. In this specification, palliative care (best supportive care) and follow-up are also treated as items of treatment. It should be noted that cancer is merely an example, and the diseases covered by this invention are not limited to cancer.

[0041] For example, in cancer treatment, the choice of treatment may depend on the stage of the cancer. Therefore, it is desirable to estimate the stage of the cancer as it develops. In other words, the method of the present invention is particularly effective when administering drug therapy, surgery, radiotherapy, palliative care (best supportive care), and observation, either alone or in combination.

[0042] (Seventh Embodiment) The patient information analysis method according to the seventh embodiment of the present invention is characterized by the selection of two patient groups: a group of patients with a good prognosis and a group of patients with a poor prognosis. As mentioned above, in order to estimate the factors that influence the outcome of treatment (good or bad), it is necessary to determine the significant difference in specific data items between two patient groups: a group with a good outcome and a group with a poor outcome. In other words, the method of the present invention is particularly effective when two patient groups are selected: a group of patients with a good prognosis and a group of patients with a poor prognosis.

[0043] (Eighth embodiment) The patient information analysis method according to the eighth embodiment of the present invention is characterized by being applied to identify factors that affect the treatment completion rate. In the medical field, continuous efforts are being made to improve treatment completion rates, and identifying the factors that influence these rates is extremely important. In other words, the method of the present invention is particularly effective when applied to identify factors that affect the completion rate of treatment.

[0044] (Ninth embodiment) As a ninth embodiment of the present invention, a patient information analysis device is described, which includes: a diagram display means that displays the order of treatment on a first axis and the type of treatment on a second axis, with the starting point of the order of treatment on the first axis and the result or intermediate state of treatment on the ending point, based on the electronic medical record data of a plurality of patients; a patient selection means that allows the selection of one or more patient groups consisting of a plurality of patients who have the same treatment in the diagram display shown by the diagram display means; and a patient information analysis means that analyzes the electronic medical record data relating to the patient group when the patient selection means has selected the patient group.

[0045] Figure 5 is a functional block diagram showing the schematic functional configuration of a patient information analysis device according to the ninth embodiment of the present invention. The patient information analysis device 500 consists of an input / output means 501 for inputting various data and outputting analysis results, a database 502 for storing various data, a display means 503 for displaying diagrams, etc., a GUI means 504 for giving instructions and making selections on the displayed diagrams, etc., an information processing means 505 for performing various analyses, calculations, and information processing, and a communication means 506 for communicating these together.

[0046] The input / output means 501, display means 503, and GUI means 504 are, for example, a tablet PC equipped with a touch display. In other words, it is possible to perform tasks such as entering medical record data, displaying dialogs, and performing operations, giving instructions, and making selections on a tablet PC, all in one integrated manner. The above method does not necessarily require a tablet PC; any device that allows for inputting and referencing medical record data, presenting and displaying analysis results, and providing instructions for operation, such as a personal computer (PC) or smartphone, is also acceptable.

[0047] Database 502 is implemented using a storage device such as an HDD (hard disk drive), and stores various data such as medical record data, date and time of implementation, and personal IDs in an interconnected format. Note that personally identifiable data may be replaced with a format that does not identify individuals during analysis processing or presentation of analysis results.

[0048] The information processing means 505 is a computational function module that realizes information processing corresponding to the patient information analysis method of the first to seventh embodiments of the present invention by having the central processing unit (hereinafter referred to as "CPU") of a computer execute program code. The communication means 506 is, for example, an internal bus of a computer, an external network such as the internet, or both an internal bus and an external network, and communicates each functional module of the patient information analysis device of the present invention to send and receive data.

[0049] Figure 6 is a flowchart illustrating the decision-making and processing flow performed by the CPU of a patient information analysis device according to the ninth embodiment of the present invention. Based on the medical record data of multiple patients, the device displays a diagram showing the order of treatment on the first axis, the type of treatment on the starting side of the treatment order on the first axis, and the treatment result or intermediate state on the ending side of the second axis. The device accepts the selection of a patient group by the user in the displayed diagram, analyzes the electronic medical record data related to the selected patient group, and displays the analysis results.

[0050] The patient information analysis device according to the present invention first displays the user interface (hereinafter referred to as "UI") of the input / output means 501 (display unit) (S601). When the user operates on the UI and specifies a database (YES in S602), the patient information analysis device of the present invention creates a diagram from that database (S603) and displays it on the input / output means 601 (display unit) (S604). When the user points to a link that is not selected by pointer operation (YES in S605), the device selects the patient group corresponding to that link as data (S606). The link corresponding to that patient group (the pointed link) is colored and displayed in a different color than before (S607). Also, when the user points to a link that was selected by pointer operation (YES in S608), the device deselects the patient group corresponding to that link as data (S609), and returns the link corresponding to that patient group (the pointed link) to the same color as when it was first displayed (S610).

[0051] When one or more links are selected and the user selects the menu item "Analyze" (YES in S611), the system performs data analysis on the selected patient group (S612) and displays the analysis results (S613). When the user selects the menu item "Finish" (YES in S614), the system hides the diagram and analysis results (S615) and terminates the processing of this flow.

[0052] Figure 7 shows a schematic configuration of a computer capable of realizing a cognitive function change prediction device according to the second embodiment of the present invention. In Figure 7, 700 is a computer, comprising a control unit 701, a storage unit 702, a peripheral device interface unit 703, an input unit 704, a display unit 705, and a communication unit 706, all connected by a bus 710. This configuration is just one example, and various configurations can be adopted as appropriate.

[0053] The control unit 701 consists of a CPU (Central Processing Unit), ROM (Read Only Memory), RAM (Random Access Memory), etc. The CPU calls programs stored in the memory unit 702, ROM, recording medium, etc., into the work memory area on RAM and executes them, drives and controls each device connected via the bus 710, and realizes the processing that a computer would perform. ROM is a non-volatile memory that holds programs and data such as the computer 700's boot program and BIOS. RAM is a volatile memory that temporarily holds programs and data loaded from the memory unit 702, ROM, recording medium, etc., and also has a work area used by the control unit 701 when performing various processing. The storage unit 702 is, for example, an HDD (hard disk drive) and stores the program executed by the control unit 701, as well as various other data.

[0054] The peripheral device interface (I / F) section 703 is a port for connecting the computer 700 to peripheral devices. The peripheral device interface section 703 consists of ports such as USB, IEEE1394, and RS-232C. The connection method to peripheral devices can be either wired or wireless. The input unit 704 has a keyboard, a mouse or other pointing device, a numeric keypad or other input device, and provides operation instructions, action instructions, data input, etc. to the computer 700. The display unit 705 is a logic circuit or device driver for displaying images, videos, etc., on a display device such as a liquid crystal panel.

[0055] The communication unit 706 has a communication control device, a communication port, etc., and is a wired or wireless communication interface that mediates communication with the network. Bus 710 is a communication path that mediates the exchange of control signals, data signals, etc., between each device.

[0056] (Other embodiments) The present invention can also be realized by supplying software (programs) that realize the functions of the embodiments described above to a system or device via a network or various storage media, and by a process in which the computer (or CPU, MPU, etc.) of that system or device reads and executes the program. Therefore, the program code installed on the computer to implement the functional processing of the present invention also constitutes an implementation of the present invention. In other words, the present invention includes the computer program itself for implementing the functional processing of the present invention. In this case, as long as it has the functionality of a program, it may take the form of object code, a program executed by an interpreter, a script or macro executed by an application program such as a web browser, or an API (Application Programming Interface). It may also be implemented by being incorporated as part of another web service (for example, a Social Networking Service (SNS)) using web programming techniques such as "Mashup".

[0057] Furthermore, the present invention does not necessarily have to consist of a single, integrated piece of hardware or software; it may also be realized by incorporating software modules into multiple pieces of hardware such as information terminals and servers, with these working together. Furthermore, the present invention is not limited to the embodiments described above, and various modifications and variations are possible without departing from the spirit and scope of the invention. In other words, the present invention is not limited by the illustrative description or the accompanying drawings. [Industrial applicability]

[0058] The patient information analysis device and patient information analysis program of the present invention can be used in electronic medical record systems for healthcare professionals and in tablet terminal devices for patients.

Claims

1. A method implemented in a computer for analyzing patient information, Based on electronic medical record data from multiple patients, the first axis displays the sequence of treatments, and the second axis displays the treatment method on the starting point side of the treatment sequence on the first axis, and the treatment result or intermediate state on the ending point side, in a diagram. In the diagram display, it is possible to select one or more patient groups consisting of multiple patients who share the same treatment, When the aforementioned patient group is selected, the electronic medical record data relating to that patient group is analyzed. A method for analyzing patient information, including the patient's information.

2. The patient information analysis method according to claim 1, further comprising, when two patient groups consisting of multiple patients with common treatments are selected in the diagram display, creating a machine learning model that predicts the classification of the two selected patient groups using the electronic medical record data of the selected patient groups.

3. The patient information analysis method according to claim 2, further comprising identifying factors that contribute to the classification of patient groups predicted by the machine learning model created above.

4. The patient information analysis method according to claim 1, further comprising determining a significant difference between the two selected patient groups with respect to at least one data item of the electronic medical record data of the selected patient groups when two patient groups consisting of multiple patients with common treatment are selected in the diagram display.

5. The method for analyzing patient information according to claim 1, wherein the treatment includes at least one selected from the group consisting of drug therapy, surgery, radiotherapy, palliative care, and observation.

6. The method for analyzing patient information according to claim 2, wherein two patient groups are selected: a group of patients with a good prognosis and a group of patients with a poor prognosis.

7. A method for analyzing patient information according to claim 3, for identifying factors that affect the treatment completion rate.

8. A patient information analysis device that analyzes patient information, A diagram display means that, based on the electronic medical record data of multiple patients, displays the order of treatment on the first axis, the type of treatment on the starting point side of the treatment order on the first axis, and the result or intermediate state of treatment on the ending point side of the second axis, In the diagram display shown by the diagram display means, a patient selection means is provided to allow the selection of one or more patient groups consisting of multiple patients who share the same treatment. A patient information analysis device characterized by having a patient information analysis means that analyzes electronic medical record data relating to a patient group when the patient group is selected by the patient selection means.

9. The patient information analysis device according to claim 8, further comprising a machine learning model creation means for creating a machine learning model that predicts the classification of the two selected patient groups when two patient groups consisting of multiple patients with common treatments are selected in the diagram display shown by the diagram display means.

10. The patient information analysis device according to claim 9, further comprising a contributing factor analysis means for analyzing factors that contribute to differentiating between the two selected patient groups using a machine learning model created by the machine learning model creation means.

11. The patient information analysis device according to claim 8, further comprising significance determination means for determining a significant difference between the two selected patient groups with respect to at least one data item of the electronic medical record data of the selected patient groups when two patient groups consisting of multiple patients with common treatment are selected in the diagram display shown by the diagram display means.

12. A program that causes a computer to function as each means in the patient information analysis device according to claim 8.

13. A computer-readable recording medium that stores a program for causing a computer to function as each means in the patient information analysis device according to claim 9.

Images