TW201426624A - Personal medical expense prediction system - Google Patents

Abstract

A personal medical expense prediction system is disclosed, comprising: a first database which includes a clinical classification table that includes clinical classification code fields, and cost fields recording medical costs corresponding each of clinical classification codes, where the clinical classification table is arranged in order basing on the cost fields; a second database which includes a prediction model for recording risk groups and corresponding risk weighting values; a user interface unit, for inputting basic information of a specific insured people and cost coefficient values applicable to all insureds; a risk weighting values inquiry unit, for classifying the specific insured people into one of the risk groups according to gender, age, history of diagnosis codes and date of diagnosis of the specific insured people, and then checking out the corresponding risk weighting values; and a prediction cost calculation unit, for calculating predicted medical expense of the specific insured people according to the risk weighting values and cost coefficient values.

Description

Personal medical expenses forecasting system

The present invention relates to a cost prediction system, and more particularly to a personal medical cost prediction system.

At present, the integrated nursing system for family physicians implemented in China, in part, uses the method of paying people to calculate the cost of virtual western medical clinics that need to be given to the community medical group. Therefore, there is a set of risk correction theory charging model suitable for China and accurate. In addition, the second-generation health insurance regulations require the establishment of a family-responsible physician system in the future for universal health insurance in China, and the payment should be based on the principle of implementation. Furthermore, the regional allocation of the total budget system for universal health insurance also needs to correct the medical needs of the regional population (reflected in future medical expenses). The above three national health insurance policies, which are currently being implemented or will be implemented in the future, are urgently needed for the estimation of personal medical expenses. Therefore, for policy makers, there needs to be a set of tools to correctly estimate personal expenses as the basis for decision-making on budget allocation, and for hospitals or community medical groups, it is necessary to estimate the medical care of the insured for one year. Costs as a budget control tool for health management. Therefore, both policy makers and medical institutions affected by the policy have strong demand for this. However, at present, there is no good personal medical cost estimation tool in China, so it is necessary to find a solution.

Accordingly, it is an object of the present invention to provide a personal medical cost prediction system.

Thus, the personal medical expenses prediction system of the present invention includes a first data a library, a second database, a user interface unit, a risk weight value query unit, and a predictive cost computing unit. The first database includes a clinical classification ranking table based on a historical diagnostic code of a predetermined number of insurance objects, wherein the clinical classification ranking table includes a clinical classification code plus a last quarter of the clinical classification. And a cost field for recording medical expenses corresponding to each clinical classification code, and the clinical classification ranking table is sorted according to the cost field. The second database includes a prediction model for recording a plurality of different risk groups and a plurality of corresponding risk weight values, wherein the risk groups are sorted according to the age, sex, and the clinical classification of each insurance object. The clinical classification code and its corresponding diagnosis date sorted by the table are constructed by a data exploration analysis method. The user interface unit is configured to input a set of basic data belonging to a specific insurance object and a set of cost coefficient values applicable to all insurance objects, wherein the basic information includes the gender of the specific insurance object, and the age of a specific year. At least one historical diagnostic code, and a corresponding diagnostic date of each historical diagnostic code. The risk weight value query unit is configured to classify the specific insurance object into the prediction model of the second database according to the gender, age, historical diagnosis code and diagnosis date of the specified specific insurance object. A risk group, and then query the corresponding risk weight value. The predicted cost calculation unit is configured to calculate the predicted medical expenses of the specific insurance object for the next year of the specific year according to the risk weight value queried by the risk weight value query unit and the cost coefficient value input by the group.

The effect of the present invention is that, for policy decision makers, it can be used as a tool for correctly estimating personal expenses as a basis for decision making for budget allocation, and For hospitals or community medical groups, it is possible to estimate the medical expenses of the insured for one year as a budget control tool for health management.

The above and other technical contents, features and advantages of the present invention will be apparent from the following detailed description of the preferred embodiments.

Before the present invention is described in detail, it is noted that in the following description, similar elements are denoted by the same reference numerals.

Referring to FIGS. 1 to 4, a preferred embodiment of the personal medical expenses prediction system of the present invention comprises a first database 1, a second database 2, a user interface unit 3, a risk weight value query unit 4, and a The fee calculation unit 5 is predicted.

The first database 1 includes a historical diagnostic code (for example, a historical diagnostic code of 1 million people in 2010) based on a predetermined non-extremely large number of insurance objects (for example, 1 million people). The clinical classification of the table 10, the so-called cost non-extreme refers to the person whose medical expenses are not greater than the highest 0.1%, about 1 million yuan. As shown in the flow chart of the construction process of the clinical classification ranking table of FIG. 2, in the preferred embodiment, the International Classification of Diseases is used for the outpatient and inpatient diagnosis of the 1 million people in the year 2010. The ICD) code, as shown in step 191, classifies the historical ICD code of 1 million people into several CCS categories by Clinical Classifications Software (CCS). For example, if the historical ICD code of 1 million people is classified by the CCS software for the ICD9 version, it can be classified into 260 CCS major categories. In addition, these CCS The clinical classification code is divided into outpatient clinical classification code and residential clinical classification code. In the embodiment of the process of constructing the clinical classification ranking table 10 of the present invention, when the clinical classification code of an outpatient clinic of an insurance object appears twice or more (including two times), the outpatient clinical condition of the insurance object is The classification code is included. As for the clinical classification code of the insured person, all of them are directly included.

Next, as shown in step 192, the total personal medical expenses (the western medical insurance outpatient service fee, the hospitalization fee, and the pharmacy cost) of each insurance object for the next year (2011) are obtained, but the insurance with the cost greater than the 99.9 percentile is excluded. Object. Then, as shown in step 193, the total personal medical expenses of all the insurance objects of the CCS clinical classification code in the same quarter are averaged to obtain the average medical expenses corresponding to each clinical classification code. Then, as shown in step 194, sorting according to the average medical expenses corresponding to the clinical classification codes of each quarter, the clinical classification ranking table 10 shown in Table 1 below as shown in FIG. 1 can be obtained. The cost field of the average medical expenses is omitted in Table 1, and in order to save the specification, only the records ranked 10, 20, 30, ..., 260 are listed, but from this simplified Table 1, It can be seen that the overall impact of the previous year's CCS on the medical expenses of the next year, for example, the average medical cost of the 260th anterior caesarean section (CCS is 189) is the lowest, so it is the least. Serious.

The second database 2 includes a prediction model 20 for recording a plurality of Risk-Adjusted Clusters (RACs) and a plurality of corresponding risk weight values, wherein the risk groups are based on the respective insurance objects. Age, gender, and clinical classification codes sorted according to the clinical classification ranking table 10 and their corresponding diagnosis dates were constructed using a data exploration analysis method. In the following In the description of the example, the Classification and Regression Tree (CART) analysis method in the data exploration analysis method will be taken as an example, and the construction process of the classification and regression tree in the embodiment will be described with reference to FIG. 3.

As shown in step 291 of FIG. 3, the ICD history diagnostic code of each insurance object can be classified into an appropriate CCS clinical classification code, and sorted according to the clinical classification sorting table 10 as shown in Table 1 to obtain each insurance. The individual's own personal CCS severity ranking. In this embodiment, the six of the highest medical expenses among all the clinical classification codes of each insurance object are selected for subsequent CART analysis.

Next, as shown in step 292, all insurance objects are classified into multi-level clinical practice standardized by the CCS clinical classification software according to the highest clinical cost code in the clinical classification code of each insurance object. The first layer of clinical classification code in the classification code. For example, if you perform this step 292 with the CCS software for ICD9 version, all insurance objects will be classified into 17 first-level sub-populations, such as circulatory diseases, infectious diseases, cancer, endocrine, metabolic diseases, etc. 17 sub-groups.

Then, as shown in step 293, the future medical expenses of the insurance target individual are used as the prediction target, and the six clinical classification codes and the corresponding diagnosis date quarters (or seasons, respectively) of each insurance object obtained in step 291 are used. Representing the impact of the clinical classification code on future expenses due to the time of disease occurrence), risk factors such as age and gender are predictors, and the CART software is used to conduct data exploration on the 17 sub-groups generated in step 292. The personal medical expenses prediction system of the present invention can be obtained to predict personal medical expenses Several RAC risk groups required for use. For example, taking the classification process of infectious diseases as an example, the tree diagram of the classified subpopulations produced is shown in Figures 4a and 4b. The last node of each branch in Figures 4a, 4b (a total of 7), which represents 7 groups of RAC risk groups for infectious diseases.

Then, as shown in step 294, the risk weight value for each risk group is equal to the average cost of each risk group divided by the average cost of all insurance object samples. For example, in the present embodiment, the grouping logic and the corresponding risk weight values of the seven risk groups generated for the infectious diseases of FIGS. 4a and 4b are W1 to W7, respectively. For example, from the leftmost first risk group, except for the most serious diagnosis of infectious diseases, the second most serious disease (CCS2) collection belongs to (4_1, 10_4, 7_2, 7_3... One of them, according to the other diseases as the basis for classification, the final number of people in this group is 21061, the average number of points in this group is 7179 points. Since the average number of points of the insurance object is 20724 when the model is constructed, the weight W1 of the risk group is 7179÷20724=0.3464.

The user interface unit 3 is presented on a user's computer or mobile device (eg, a tablet computer, etc.) for the user (eg, health manager, medical administrator, physician, etc.) to enter a set of basic items belonging to a specific insurance object. Information and a set of cost factor values that apply to all insured objects.

The basic information of the group includes the gender of the specific insurance object, the age of a specific year (for example, 2011), at least one historical diagnosis code, and the corresponding diagnosis date of each historical diagnosis code.

The set of cost factor values includes the average cost per person A for that particular year and the annual growth rate B for the next year (eg, 2012) for that particular year.

After the user completes the input of the basic data and the cost coefficient value by using the user interface unit 3, the risk weight value query unit 4 first selects the specific information according to the gender, age, historical diagnosis code and diagnosis date of the specified specific insurance object. The insurance object is classified into one of the risk groups in the prediction model 20 of the second database 2, and the corresponding risk weight value is queried from the prediction model 20. For example, when the risk weight value inquiry unit 4 returns the specific insurance object according to the age of the specific insurance object (for example, 2011), the historical diagnosis code, and the corresponding diagnosis date of each historical diagnosis code input by the user. Figure 4a, 4b Infectious disease CART classification sub-population tree diagram The next-most leftmost risk group indicates the predicted risk of the specific insurance object in the next year (ie 2012) of the specific year The weight value is W1.

Then, the predicted fee calculation unit 5 calculates the predicted medical cost of the next year of the specific insurance object for the specific year = the average per person fee A × (1 + next year growth rate B) × the risk weight value. For example, when the risk weight value query unit 4 queries a specific insurance object, the predicted risk weight value in 2012 is W1 and the average per capita health insurance medical cost in 2011 is A=19900 yuan and the annual growth rate in 2012B. When the value is 3.92%, the predicted cost calculation unit 5 calculates that the predicted medical expenses of the specific insurance object in 2012 is equal to 7163 yuan.

In summary, the personal medical expenses prediction software system of the present invention uses a clinical classification ranking table in the first database, a prediction model in the second database, a user interface unit, a risk weight value query unit, and a prediction cost calculation unit. Cooperative operation between program components, which can be used by policy decision makers as correct A tool for estimating personal expenses as a basis for decision-making on budget allocation, and can be used by hospitals or community medical groups to estimate the annual medical expenses of insurance applicants, as a budget control tool for health management, so it can be achieved. The object of the invention.

The above is only the preferred embodiment of the present invention, and the scope of the invention is not limited thereto, that is, the simple equivalent changes and modifications made by the scope of the invention and the description of the invention are All remain within the scope of the invention patent.

1‧‧‧First database

10‧‧‧Clinical Classification Sorting Table

191~194‧‧‧Steps

2‧‧‧Second database

20‧‧‧ Forecast Model

291~294‧‧‧Steps

3‧‧‧User interface unit

4‧‧‧Risk weight value query unit

5‧‧‧Predicted costing unit

1 is a system block diagram illustrating a preferred embodiment of the personal medical cost prediction system of the present invention; and FIG. 2 is a flow chart illustrating the construction of a clinical classification ranking table in the first database of the preferred embodiment of the present invention; 3 is a flow chart illustrating a process flow for constructing a predictive model in a second database in a preferred embodiment of the present invention; and FIGS. 4a and 4b are schematic views illustrating CART in a preferred embodiment of the present invention; Infectious disease classification subpopulation tree diagram.

1‧‧‧First database

10‧‧‧Clinical Classification Sorting Table

2‧‧‧Second database

20‧‧‧ Forecast Model

3‧‧‧User interface unit

4‧‧‧Risk weight value query unit

5‧‧‧Predicted costing unit

Claims (10)

A personal medical cost prediction system includes: a first database comprising a clinical classification ranking table based on a historical diagnostic code of a predetermined number of insurance objects, wherein the clinical classification ranking table includes a clinical classification code field And a cost field for recording the average medical expenses corresponding to each clinical classification code, and the clinical classification ranking table is sorted according to the cost field; a second database includes a record of multiple different risk groups And a plurality of corresponding risk weight value prediction models, wherein the risk groups are based on the age, sex, and the clinical classification code sorted according to the clinical classification ranking table and corresponding diagnostic dates, using a data Constructed by the exploration analysis method; a user interface unit for inputting a set of basic data belonging to a specific insurance object and a set of cost coefficient values applicable to all insurance objects, wherein the basic information includes the specific insurance object Gender, age for a specific year, at least one historical diagnostic code, and corresponding diagnostic date for each historical diagnostic code a risk weight value query unit for classifying the specific insurance object into the prediction model of the second database according to the gender, age, historical diagnosis code and diagnosis date of the specified specific insurance object a risk group, and then querying the corresponding risk weight value; and a predicted cost computing unit, configured to calculate the risk weight value queried by the risk weight value query unit and the cost coefficient value input by the group The specific insurance object is in the next year of the specific year Predict medical expenses. According to the personal medical expenses prediction system described in claim 1, wherein the clinical classification codes include an outpatient clinical classification code and a residential clinical classification code, and in the process of constructing the clinical classification ranking table, one of them The specific outpatient clinical classification code of the insurance object is included when the specific outpatient clinical classification code of the insurance object appears twice or more. According to the personal medical expenses prediction system described in the first application scope of the patent application, in the process of constructing the clinical classification ranking table of the first database, all clinical classification codes and each insurance of each insurance object are obtained first. The total personal medical expenses of the subject, and then the average personal medical expenses of all the insurance objects of the same clinical classification code are averaged to obtain the average medical expenses corresponding to each clinical classification code. According to the personal medical expenses prediction system described in claim 1, wherein in the process of constructing the prediction model by using the data exploration analysis method, each of the insurance objects is sorted based on the clinical classification ranking table. The month of the diagnosis date of the clinical classification code is divided by 12 to represent the weight of the clinical classification code for future impact. According to the personal medical expenses prediction system described in claim 4, wherein the historical diagnosis code is an international disease classification code, and the clinical classification code in the first database is by using a clinical classification software. International disease classification codes are classified. According to the personal medical expenses prediction system described in claim 5, wherein the risk group in the prediction model is selected by first selecting the highest average medical cost among all clinical classification codes of each insurance object. Use the clinical classification code, and then classify all the insurance objects into the first-level clinical classification in the multi-level clinical classification code standardized by the clinical classification software according to the clinical classification code with the highest average medical cost in the clinical classification code of each insurance object. The code is then constructed using the data exploration analysis method according to the six clinical classification codes with the highest age, sex, and average medical expenses of each insurance object and the corresponding diagnosis date quarter. According to the personal medical expenses prediction system described in claim 1, wherein the data exploration analysis method is a classification and regression tree analysis method. According to the personal medical expenses prediction system described in claim 1, wherein the corresponding risk weight value of each risk group is equal to the average cost of each risk group divided by the average cost of all the insurance objects of the predetermined number of people. The personal medical expense prediction system of claim 8 wherein the set of cost factor values includes an average per person fee for the particular year and an annual growth rate for the next year of the particular year. The personal medical expenses prediction system according to claim 9, wherein the predicted fee calculation unit calculates a predicted medical cost of the specific insurance subject for the next year of the specific year = an average cost per person for the specific year × The annual growth rate of universal health insurance costs × the risk weight value of the risk group.