Intelligent disability grade evaluation method based on big data analysis

Abstract

The invention relates to an intelligent disability grade evaluation method based on big data analysis, and the method comprises the following steps: data collection: combining medical data, and collecting all patient data and partial disability patient data in a time period; factor screening: selecting disease factors and fusing the disease factors into characteristic factors such as patient age and admission frequency; performing data cleaning treatment, and establishing a model by using a medical data evaluation principle and the obtained medical insurance settlement list data; data analysis: introducing a logistic regression algorithm to train and verify the model; and analyzing a result, and introducing an index to evaluate the performance of the model. According to the method, the modern information technology is creatively utilized, evaluation suggestions with scientific basis are provided for a traditional disability grade evaluation conclusion, and the evaluation suggestions comprise intermediate results of historical disease course records and immediate activity ability verification. The problem that disability level evaluation more depends on the management ability, professional accomplishment, professional operation and the like of evaluation institutions is solved, and effective application of the multiplier effect of medical insurance data is accelerated and stimulated.

Description

Technical Field

[0001] The present invention relates to the field of quality management and health assessment, and in particular to an intelligent evaluation method for disability levels based on big data analysis. Background Art

[0002] Disabled persons refer to those who have partially or completely lost their living or social abilities due to old age, illness, disability, etc., and must rely on others for assistance or completely rely on others to complete daily activities such as eating, dressing, getting in and out of bed, going to the toilet, walking around indoors, and bathing. Depending on the degree of disability, they can be divided into mild disability, moderate disability, and severe disability.

[0003] Traditional disability level assessment relies on assessment service personnel based on assessment standards. The assessment principle is basically based on the classification of ADL, and comprehensively assesses the cognitive, visual, and communication abilities of the assessment object. The current assessment method is mainly manual assessment, which relies on professional experience to make judgments. The on-site assessment of the assessor strictly follows this standard and draws the final assessment conclusion after certain calculations. The overall subjectivity is relatively strong, and the participation of data elements is low, resulting in uneven quality of disability level assessment, making it difficult to obtain objective and fair assessment results, which brings many unstable factors to the subsequent supervision and payment of benefits. Summary of the invention

[0004] In view of the above-mentioned shortcomings of the prior art, an object of the present invention is to provide an intelligent evaluation method for disability level based on big data analysis to solve one or more problems in the prior art.

[0005] To achieve the above object, the technical solution of the present invention is as follows:

[0006] The intelligent evaluation method of disability level based on big data analysis includes the following steps:

[0007] Data collection, combining medical data and collecting all patient data within a time period and some disabled patient data;

[0008] Factor screening: selecting disease factors and integrating characteristic factors such as patient age and number of hospitalizations;

[0009] Data cleaning and management, using medical data evaluation principles and acquired medical insurance settlement list data to build models;

[0010] Data analysis, introducing logistic regression algorithm to train and verify the model;

[0011] Result analysis,introduces indicators to evaluate the performance of the model.

[0012] Furthermore, the disease factors include nervous system lesions, muscular system injuries, skeletal system injuries, combined factors and functional impairment manifestations.

[0013] Furthermore, the data cleaning management includes the following steps:

[0014] Combine the personnel codes and hospitalization data of disabled persons to match the names of diseases and disease codes before the persons became disabled;

[0015] Clean disease codes and illness codes in conjunction with the coding manual;

[0016] Combined deduplication yields the number of hospitalizations of disabled persons during the time period;

[0017] Calculate the statistical distribution of symptoms of disabled people and derive the disability factor based on the medical data evaluation principle;

[0018] Filter and mark some valid disabled and non-disabled personnel data.

[0019] Furthermore, the introduction of the logistic regression algorithm comprises the following steps:

[0020] Select the model;

[0021] Construct prediction function;

[0022] Determine the loss function;

[0023] Optimize the loss function.

[0024] Furthermore, the prediction function is obtained by performing nonlinear transformation based on the sigmoid function, and the sigmoid function is shown in the following formula:

[0025]

[0026] Where: z is the training data.

[0027] Furthermore, the prediction function is shown in the following formula:

[0028]

[0029] In the formula: x is the independent variable and θ is the coefficient of the equation.

[0030] Furthermore, the loss function is determined by taking the log-likelihood function as the optimization target and deriving it based on the maximum likelihood estimation. The loss function is expressed by combining the Cost function and the J function, as shown in the following formula:

[0031]

[0032]

[0033] In the formula: m represents the sample size, θ is the equation coefficient, x is the independent variable, and y is the dependent variable.

[0034] Furthermore, the optimization loss function is implemented by a gradient descent algorithm to find the optimal equation coefficients. The optimization algorithm is shown in the following formula:

[0035]

[0036] In the formula: α is the learning rate, m represents the number of samples, θ is the equation coefficient, x is the independent variable, y is the dependent variable, Represents x i The jth feature of .

[0037] Furthermore, after the optimization algorithm and combined with regularization, the expression of the optimized loss function is shown as follows:

[0038]

[0039] In the formula: m represents the number of samples, n is the number of features of each sample, θ is the equation coefficient, x is the independent variable, y is the dependent variable, λ is the regularization coefficient, θ j Represents the coefficient of the j-th feature.

[0040] Furthermore, the result analysis is obtained by calculating the confusion matrix, including accuracy, precision, recall, F1-score, ROC curve and AUC value; the values ​​of the confusion matrix include true positives, true negatives, false positives and false negatives.

[0041] Compared with the prior art, the beneficial technical effects of the present invention are as follows:

[0042] The present invention innovatively utilizes modern information technology, fully explores the application value of data elements, comprehensively analyzes multi-source data, efficiently identifies high-risk groups for disability, and gives a scientific evaluation of their ability to take care of themselves in daily life. At the same time, it can predict the nursing service needs of disabled people. The invention breaks through the current traditional disability level assessment method and introduces another new solution for the quality control of disability level assessment. It provides scientifically based evaluation suggestions for the conclusions of traditional disability level assessments, including intermediate results verified by historical medical records and immediate activity capabilities. It solves the problem that disability level assessments rely more on the management capabilities, professional qualities, and professional ethics of assessment agencies. It also solves the problems of rent-seeking, human interference, and disguised insurance fraud in assessment services, and accelerates the effective use of the multiplier effect of medical insurance data. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 The overall process diagram of the intelligent evaluation method for disability level based on big data analysis in an embodiment of the present invention is shown.

[0044] Figure 2A schematic diagram of the data cleaning process of the method for intelligent evaluation of disability levels based on big data analysis in an embodiment of the present invention is shown.

[0045] Figure 3 A schematic diagram of the data analysis process of the method for intelligent evaluation of disability levels based on big data analysis according to an embodiment of the present invention is shown.

[0046] Figure 4 A schematic diagram of a sigmoid function image of a method for intelligently evaluating disability levels based on big data analysis according to an embodiment of the present invention is shown.

[0047] Figure 5 A schematic diagram showing the regularized fitting degree of the intelligent evaluation method for disability level based on big data analysis according to an embodiment of the present invention is shown.

[0048] Figure 6 A schematic diagram showing the model verification parameter results of the disability level intelligent evaluation method based on big data analysis according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical scheme and advantages of the present invention clearer, the intelligent evaluation method of disability level based on big data analysis proposed by the present invention is further described in detail below in combination with the accompanying drawings and specific embodiments. According to the following description, the advantages and features of the present invention will be clearer. It should be noted that the accompanying drawings adopt a very simplified form and use non-precise proportions, which are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention. In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, please refer to the accompanying drawings. It should be noted that the structure, proportion, size, etc. illustrated in the accompanying drawings of this specification are only used to match the content disclosed in the specification, so that people familiar with this technology can understand and read, and are not used to limit the limiting conditions for the implementation of the present invention, so it has no technical substantive significance. Any modification of the structure, change of the proportional relationship or adjustment of the size, without affecting the effect that the present invention can produce and the purpose that can be achieved, should still fall within the scope of the technical content disclosed by the present invention.

[0050] See also Figure 1 , the intelligent evaluation method of disability level based on big data analysis includes the following steps:

[0051] Step 1: Data collection, combining medical data and collecting all patient data within the time period and some disabled patient data.

[0052] The time period data is collected from the provincial front-end machine. It is assumed that the hospitalization data of the whole province from 21 to 23 years is collected, totaling more than 28 million records. The fields include the patient's ID number, personnel number, admission time, discharge time, disease name, disease code, etc. In addition, more than 3,000 disabled personnel data are also collected, and the fields include personnel number, ID card, date of birth, date of disability, etc.

[0053] Step 2: Factor screening: select disease factors and integrate characteristic factors such as patient age and number of hospitalizations.

[0054] The disease factors include nervous system lesions, muscular system injuries, skeletal system injuries, combined factors and functional impairment manifestations.

[0055] Furthermore, the nervous system lesions include neuronal degeneration and death, demyelination, inflammatory response and nerve regeneration disorder.

[0056] Furthermore, the muscle system damage includes muscle atrophy and degeneration and muscle weakness and spasticity.

[0057] Furthermore, the skeletal system injuries include fractures and joint injuries, structural damage, pain and inflammation, and joint deformity and stiffness.

[0058] Furthermore, the comprehensive factors include motor coordination and balance disorders, pain and fatigue, and secondary complications.

[0059] Furthermore, the functional impairment manifestations include motor weakness and paralysis, motor coordination disorders, gait abnormalities and sensory disturbances.

[0060] In addition to the above-mentioned disease factors, characteristic factors such as the patient's age, number of hospitalizations, and severity of the disease are also incorporated to prepare for building the model.

[0061] Step 3: Data cleaning and management, using medical data evaluation principles and the obtained medical insurance settlement list data to build a model.

[0062] Please continue reading Figure 2 Furthermore, the data cleaning management includes the following steps:

[0063] Step 3.1: Combine the personnel code and the hospitalization data of disabled persons to match the disease name and disease code before the person became disabled.

[0064] With the personnel code as the primary key, the hospitalization data is associated with the list of disabled persons to match the disease name and disease code of each disabled person before disability.

[0065] Step 3.2: Clean the disease codes and illness codes in conjunction with the coding manual.

[0066] Clean the disease codes and disease codes according to the national standard document: Medical ICD-9-CM Coding Manual.

[0067] Step 3.3: Combine and remove duplicates to obtain the number of hospitalizations of disabled persons within a time period.

[0068] Based on the personnel number and admission time in the hospitalization data of disabled persons, duplicate removal was performed and the number of hospitalizations of disabled persons in the 21 to 23 years before disability was obtained as the disability factor characteristic.

[0069] Step 3.4: Calculate the statistical distribution of symptoms of disabled persons and derive the disability factor based on medical data evaluation principles.

[0070] The statistical distribution of all diseases of disabled persons is calculated and combined with the medical data evaluation principle of "disability factor evaluation", the 100 diseases with the greatest correlation are obtained as disability factors, and finally the patient's age is added as the final disability factor.

[0071] Step 3.5: Filter and mark some valid disabled personnel data and non-disabled personnel data.

[0072] According to the above process, a total of 1,935 valid disabled personnel data were screened out and labeled as positive. At the same time, 10,000 non-disabled personnel data were randomly selected from the hospitalization information table according to the above characteristic conditions as the negative class for machine training.

[0073] Step 4: Data analysis, introducing logistic regression algorithm to train and verify the model.

[0074] The prepared positive data and negative data are put into the logistic regression algorithm for training, specifically according to an 80% ratio, that is, 1548 disabled persons' data are randomly selected from 1935 disabled persons, and 8000 non-disabled persons' data are randomly selected from 10000 non-disabled persons' data to form a training set, and the remaining 387 disabled persons' data and 2000 non-disabled persons' data are used as test sets to verify the model.

[0075] This solution mainly uses the logistic regression algorithm, which is a statistical method for solving binary classification problems, that is, 0 or 1. Logistic regression classifies by calculating the probability of an event. Its output value is a probability value, and then this probability value is compared with a preset threshold, usually 0.5, to complete the classification. The logistic regression algorithm has certain advantages in the discrimination and prediction of qualitative and semi-quantitative data. Especially when applied to discriminant analysis, it shows a high accuracy. The algorithm has the advantages of simple implementation, easy to understand and implement, strong interpretability, low computational cost, fast speed, and low storage resources.

[0076] Further, the introduction of the logistic regression algorithm, please refer to Figure 3 , including the following steps:

[0077] Step 4.1: Select the model.

[0078] The logistic regression model assumes that the output variable, i.e. the target variable, is binomially distributed, while the input variables can be continuous or discrete. The goal of logistic regression is to find the best coefficients to maximize the predictive power of the model.

[0079] Step 4.2: Construct a prediction function.

[0080] Logistic regression uses the sigmoid function, that is, the result of Y is substituted into a nonlinear transformation of the Sigmoid function to obtain a number S in the closed interval from 0 to 1. S is a probability value. The probability threshold is set to 0.5. When S is greater than 0.5, it can be regarded as a positive sample, and when it is less than 0.5, it can be regarded as a negative sample, and then classified.

[0081] Furthermore, the prediction function described above is obtained by performing nonlinear transformation based on the sigmoid function, and the sigmoid function is shown in the following formula 1:

[0082]

[0083] Where: z is the training data.

[0084] For further information, please refer to Figure 4 , for the case of linear boundaries, the boundary form is shown in Equation 2:

[0085]

[0086] Where: z is the training data, θ is the equation coefficient.

[0087] Among them, the training data is a vector as shown in the above formula 2, and the optimal parameters are shown in the following formula 3:

[0088] θ =[θ0, θ1,θ2,…,θ n ] T (3)

[0089] Further, according to the above formulas 2 and 3, the prediction function is obtained as shown in the following formula 4:

[0090]

[0091] In the formula: x is the independent variable and θ is the coefficient of the equation.

[0092] The value obtained from the function in equation 4 above represents the probability of the result being 1, so the probabilities of the classification result of input x being category 1 and category 0 are respectively as shown in equation 5 below:

[0093]

[0094] Step 4.3: Determine the loss function.

[0095] Logistic regression uses the log-likelihood loss function as the optimization objective. This loss function measures the difference between the probability distribution predicted by the model and the true probability distribution. The constructed loss function includes the Cost function and the J function. The loss function is expressed by combining the Cost function and the J function. It contains m samples and each sample has n features, all of which are derived based on the maximum likelihood estimation.

[0096] The Cost function is shown in Formula 6 below:

[0097]

[0098] In the formula, the value of y is obtained by the above formula 5.

[0099] The J function is shown in Formula 7 below:

[0100]

[0101] In the formula: m represents the sample size, θ is the equation coefficient, x is the independent variable, and y is the dependent variable.

[0102] Step 4.4: Optimize the loss function.

[0103] By minimizing the loss function, the best coefficients can be found. This is usually achieved through the gradient descent algorithm or more advanced optimization algorithms such as Newton's method, quasi-Newton's method, etc. In each iteration, the gradient of the loss function of the current parameter is calculated, and the parameters are updated in the opposite direction of the gradient so that the loss function value continues to decrease. The gradient descent method solves the minimum value, and the optimization algorithm is shown in the following formula 8:

[0104]

[0105] In the formula: α is the learning rate, m represents the number of samples, θ is the equation coefficient, x is the independent variable, y is the dependent variable, Represents x i The jth feature of .

[0106] Furthermore, python is used to read the logistic regression algorithm package, set up the logistic regression, and regularize the case data. Regularization is the implementation of the structural risk minimization strategy, which is to add a regularization term or penalty term to the empirical risk. The regularization term is generally a monotonically increasing function of the model complexity. The more complex the model, the larger the regularization term. The regularization term can take different forms. In the regression problem, the square loss is taken, which is the L2 norm of the parameter, or the L1 norm can be taken. When taking the square loss, the expression of the optimized loss function obtained after the optimization algorithm and combined with regularization is shown in Equation 9 below:

[0107]

[0108] In the formula: m represents the number of samples, n is the number of features of each sample, θ is the equation coefficient, x is the independent variable, y is the dependent variable, λ is the regularization coefficient, θ j Represents the coefficient of the j-th feature.

[0109] Regularization mainly solves the problem of overfitting. Overfitting means overfitting the training data, which increases the complexity of the model, reduces its ability to predict unknown data, and reduces its ability to predict unknown data. Figure 5 As shown, the left picture is underfitting, the middle picture is appropriate fitting, and the right picture is overfitting.

[0110] Preferably, in this case, the regularization setting is L2, the regularization strength defaults to 1, and the logistic regression loss function optimization method selects 'lbfgs'. ​​The optimal logistic regression equation is generated by iterating 10,000 times using the training set data, and then the model is used to predict the remaining test set. The model will output a probability value, and then according to the preset threshold, which is 0.5 in this embodiment, it is converted into a binary classification result.

[0111] Step 5: Result analysis: introduce indicators to evaluate the performance of the model.

[0112] Furthermore, the result analysis is obtained by calculating the confusion matrix, including accuracy, precision, recall, F1-score, ROC curve and AUC value. The values ​​of the confusion matrix include true positive, true negative, false positive and false negative. The confusion matrix is ​​shown in Table 1 below:

[0113] Actual Category Predicted as positive class Predicted as negative class Positive Category True Positive (TP) False Negative (FN) Negative False Positive (FP) True Negative (TN)

[0114] True Positives (TP) represent the number of samples that are actually positive and predicted to be positive.

[0115] True negatives (TN) represent the number of samples that are actually negative and predicted to be negative.

[0116] False positives (FP) represent the number of samples that are actually negative but predicted to be positive (also known as type I error).

[0117] False negatives (FN) represent the number of samples that are actually positive but predicted to be negative (also known as type II error).

[0118] Furthermore, the accuracy rate refers to the ratio of the number of samples correctly classified by the classifier to the total number of samples. The higher the accuracy rate, the better the performance of the classifier. It is calculated by (TP+TN) / (TP+TN+FP+FN).

[0119] Furthermore, precision refers to the ratio of the number of samples correctly predicted as positive by the classifier to the number of samples predicted as positive. The higher the precision, the lower the misjudgment rate of the classifier. It is calculated by TP / (TP+FP).

[0120] Furthermore, the recall rate refers to the ratio of the number of samples correctly predicted as positive examples by the classifier to the number of all true positive examples. The higher the recall rate, the stronger the ability of the classifier to correctly predict positive examples. It is calculated by TP / (TP+FN).

[0121] Furthermore, F1-score is the weighted harmonic mean of precision and recall, and is an indicator for evaluating the overall performance of a classifier. It is calculated by 2*precision*recall / (precision+recall).

[0122] Furthermore, the ROC curve is a curve that measures the performance of a binary classification model, and the AUC is the area under the ROC curve. The larger the AUC, the better the model performance.

[0123] Specifically, the case model validation parameters are examined through indicators, and the classification_report module in sklearn.metrics of Python is used to obtain specific results such as Figure 6 As shown: the overall accuracy of the model is about 87%, the precision of disabled persons is about 72%, the recall rate (hit) of disabled persons is 32%, the F1score is 0.45, and the ROC curve score is about 0.65.

[0124] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0125] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. An intelligent evaluation method for disability level based on big data analysis, characterized by: The following steps are involved: Data collection, combining medical data and collecting all patient data within a time period and some disabled patient data; Factor screening: selecting disease factors and integrating characteristic factors such as patient age and number of hospitalizations; Data cleaning and management, using medical data evaluation principles and acquired medical insurance settlement list data to build models; Data analysis, introducing logistic regression algorithm to train and verify the model; Result analysis,introduces indicators to evaluate the performance of the model.

2. The intelligent evaluation method for disability level based on big data analysis according to claim 1, characterized in that: The disease factors include nervous system lesions, muscular system injuries, skeletal system injuries, combined factors and functional impairment manifestations.

3. The intelligent evaluation method for disability level based on big data analysis according to claim 1, characterized in that: The data cleaning management includes the following steps: Combine personnel codes and hospitalization data of disabled persons to match the names of diseases and disease codes before the persons become disabled; clean disease codes and disease codes in combination with the coding manual; Combined deduplication yields the number of hospitalizations of disabled persons during the time period; Calculate the statistical distribution of symptoms of disabled people and derive the disability factor based on the medical data evaluation principle; Filter and mark some valid disabled and non-disabled personnel data.

4. The method for intelligent evaluation of disability level based on big data analysis according to claim 1, characterized in that: The introduction of the logistic regression algorithm comprises the following steps: Select the model; Construct prediction function; Determine the loss function; Optimize the loss function.

5. The intelligent evaluation method for disability level based on big data analysis according to claim 4 is characterized in that: The prediction function is obtained by performing nonlinear transformation based on the sigmoid function, and the sigmoid function is shown in the following formula: Where: z is the training data.

6. The method for intelligent evaluation of disability level based on big data analysis according to claim 5, characterized in that: The prediction function is shown below: In the formula: x is the independent variable and θ is the coefficient of the equation.

7. The intelligent evaluation method for disability level based on big data analysis according to claim 6, characterized in that: The loss function is determined by taking the log-likelihood function as the optimization target and deriving it based on the maximum likelihood estimation. The loss function is expressed by combining the Cost function and the J function, as shown in the following formula: In the formula: m represents the sample size, θ is the equation coefficient, x is the independent variable, and y is the dependent variable.

8. The method for intelligent evaluation of disability level based on big data analysis according to claim 7, characterized in that: The optimization loss function is implemented by a gradient descent algorithm to find the best equation coefficients. The optimization algorithm is shown in the following formula: In the formula: α is the learning rate, m represents the number of samples, θ is the equation coefficient, x is the independent variable, y is the dependent variable, Represents x i The jth feature of .

9. The method for intelligent evaluation of disability level based on big data analysis according to claim 8, characterized in that: After the optimization algorithm and combined with regularization, the expression of the optimized loss function is shown as follows: In the formula: m represents the number of samples, n is the number of features of each sample, θ is the equation coefficient, x is the independent variable, y is the dependent variable, λ is the regularization coefficient, θ j Represents the coefficient of the j-th feature.

10. The intelligent evaluation method for disability level based on big data analysis according to claim 1, characterized in that: The result analysis is obtained by calculating the confusion matrix, including accuracy, precision, recall, F1-score, ROC curve and AUC value; the values ​​of the confusion matrix include true positives, true negatives, false positives and false negatives.

Images