Evaluation model training method and system for nursing human resource allocation model based on machine learning

Abstract

An evaluation model training method of a nursing human resource allocation model based on machine learning comprises the following steps: selecting DRGs information and nursing information of an observation individual, and extracting evaluation indexes to obtain a plurality of evaluation indexes; performing single-factor analysis on the plurality of evaluation indexes, and checking the single-factor correlation between each evaluation index and the patient nursing difficulty score; performing multi-factor analysis on the plurality of evaluation indexes, checking the multi-factor correlation between each evaluation index and the patient nursing difficulty score, and analyzing the prediction performance of a multi-factor model; and correcting the regression coefficient of the multi-factor model, so that all evaluation indexes incorporated into the multi-factor model have influences on the nursing difficulty of the patient, and obtaining a nursing human resource allocation model evaluation model capable of calculating the score of the nursing difficulty of the patient. The nursing difficulty score of the inpatient can be objectively and automatically evaluated in real time, the service demand is evaluated, and the clinical nursing human resource evaluation is changed from artificial curing empirical judgment to artificial intelligence objective datamation analysis.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical big data information processing, and specifically relates to a nursing human resource configuration model training method and system based on machine learning. Background Art

[0002] In recent years, the National Nursing Development Plan (2016-2020) proposed that by 2020, the ratio of total nurses in tertiary general hospitals / some tertiary specialized hospitals to actual open beds should not be less than 0.8:1, and the ratio of total nurses in wards to actual open beds should not be less than 0.6:1. However, the current bed-to-nurse ratio in hospitals at all levels in various regions of my country is quite different from the national reference standard and foreign industry standards. Some hospitals classify patients or wards and allocate nursing manpower by modifying the foreign "patient classification system (PCS)". The evaluation indicators are complicated and varied, and a large number of indicators have low correlation with patient nursing difficulty and classification. There is a lack of reference basis for evaluating patient nursing difficulty and classification, and data redundancy can easily lead to deviations in the evaluation model's assessment of patient nursing difficulty and required nursing manpower, resulting in unreasonable allocation of nursing difficulty and nursing resources. Summary of the invention

[0003] In response to the above technical problems, we rely on multiple medical specialties to conduct comprehensive evaluation and multi-dimensional nursing resource allocation assessment, and build an evaluation index system / model based on DRGs, evaluation indicators and nursing difficulty scores with a higher correlation based on the center's information platform and combined with machine learning ideas. This improves the evaluation methods of clinical nursing service needs and patient nursing difficulty, improves the effectiveness and accuracy of nursing manpower allocation, thereby improving the level of clinical nursing management and improving the quality and efficiency of nursing work.

[0004] The first aspect of the present invention provides: a method for training an evaluation model of a nursing human resource allocation model based on machine learning, the method comprising:

[0005] Select the DRGs information and nursing information of the observed individuals, extract the evaluation indicators, and obtain several evaluation indicators;

[0006] A univariate analysis was performed on several of the evaluation indicators to test the univariate correlation between each evaluation indicator and the patient nursing difficulty score;

[0007] Conduct multivariate analysis on several of the evaluation indicators, test the multivariate correlation between each evaluation indicator and the patient care difficulty score, and analyze the predictive performance of the multivariate model;

[0008] The regression coefficient of the multi-factor model is corrected so that all evaluation indicators included in the multi-factor model have an impact on the patient's nursing difficulty, and a nursing human resource allocation model evaluation model that can calculate the patient's nursing difficulty score is obtained.

[0009] The second aspect of the present invention provides: an evaluation system for a nursing human resource allocation model based on machine learning, the system comprising at least one processor; and a memory storing instructions, which, when executed by at least one processor, implement the steps of the above-mentioned method.

[0010] The third aspect of the present invention provides: a computer-readable storage medium of an evaluation model of a nursing human resource allocation model based on machine learning, on which a computer program / instruction is stored, and the computer program / instruction implements the steps of the above method when executed by a processor.

[0011] The fourth aspect of the present invention provides: a computer program product of an evaluation model of a nursing human resource allocation model based on machine learning, including a computer program / instructions, which implement the steps of the above method when executed by a processor.

[0012] The fifth aspect of the present invention provides: a network system for an evaluation model of a nursing human resource allocation model based on machine learning, comprising: a first server and a second server, wherein the first server or the second server is capable of executing the steps of the above method.

[0013] The beneficial effects of the present invention are: by constructing an evaluation index system for the allocation of nursing human resources, the nursing service needs of hospitalized patients can be evaluated objectively, automatically and in real time, and the clinical nursing human resource evaluation is transformed from manual, rigid, empirical judgment to artificial intelligence-based objective data analysis. The existing patient classification system DRGs is linked to the difficulty of patient care. The evaluation model (or evaluation system) uses strongly correlated evaluation indicators, which requires a smaller amount of data to be collected, and the model is more efficient, while also placing less pressure on patient monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 Flowchart of the evaluation model training method according to an embodiment of the present invention;

[0015] Figure 2 Flow chart of single factor analysis of evaluation model according to the embodiment of the present invention;

[0016] Figure 3 Flow chart of multi-factor analysis of the evaluation model of the embodiment of the present invention;

[0017] Figure 4 A flowchart of the calculation of the prediction mean square error of the multi-factor analysis of the evaluation model according to the embodiment of the present invention;

[0018] Figure 5 A flow chart of regression correction for multi-factor analysis of the evaluation model according to an embodiment of the present invention;

[0019] Figure 6 Evaluation index extraction flow chart of the embodiment of the present invention;

[0020] Figure 7 The embodiment of the present invention selects the observed individual and the nursing information flow chart;

[0021] Figure 8 A schematic diagram of the structure of an evaluation system according to an embodiment of the present invention;

[0022] Fig. 9 Schematic diagram of the processor module structure of the evaluation system of the embodiment of the present invention Figure 1 ;

[0023] Fig.10 Schematic diagram of the processor module structure of the evaluation system of the embodiment of the present invention Figure 2 ;

[0024] Fig.11 Flowchart of a method for constructing a nursing human resources configuration model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following examples further illustrate the content of the present invention, but should not be construed as limiting the present invention. Without departing from the spirit and substance of the present invention, modifications or substitutions made to the methods, steps or conditions of the present invention all fall within the scope of the present invention.

[0026] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0027] my country's existing medical technology platform relies on DRGs (Diagnosis Related Groups) to form a comprehensive system of medical insurance amount, medical care performance, and medical evaluation. In clinical care, patients who are divided into the same DRGs group according to the medical difficulty evaluation often have similar symptoms and similar consumption of nursing resources. The use of existing DRGs to analyze nursing resources and nursing difficulty has stronger universality. DRGs-related information includes comprehensive information such as the patient's diagnosis. Using the DRGs analysis results as a reference factor for the human resource allocation model can simplify the total amount of patient information collection and processing; at the same time, it can further optimize human resource allocation.

[0028] Given the limited existing human and medical resources, it is somewhat difficult to fully meet the national standards. By optimizing the existing human resource allocation structure, we can use the ward as a nursing unit and take the total / average difficulty score of a ward as a baseline, which can enable more flexible adaptation to human resource allocation.

[0029] In some embodiments of the present invention, Figure 1-10 As shown, a method for constructing an evaluation model of a nursing human resource configuration model based on machine learning includes:

[0030] Select the DRGs information and nursing information of the observed individuals, extract the evaluation indicators, and obtain several evaluation indicators;

[0031] A univariate analysis was performed on several of the evaluation indicators to test the univariate correlation between each evaluation indicator and the patient nursing difficulty score;

[0032] Conduct multivariate analysis on several of the evaluation indicators, test the multivariate correlation between each evaluation indicator and the patient care difficulty score, and analyze the predictive performance of the multivariate model;

[0033] The regression coefficient of the multi-factor model is corrected so that all evaluation indicators included in the multi-factor model have an impact on the patient's nursing difficulty, and a nursing human resource allocation model evaluation model that can calculate the patient's nursing difficulty score is obtained.

[0034] In these embodiments:

[0035] The evaluation model of the nursing human resource allocation model based on machine learning is used to provide corresponding nursing difficulty scores according to the patient's information. The nursing difficulty scores can serve as an important reference for subsequent human resource allocation.

[0036] DRGs is essentially a case combination classification scheme, that is, a system that divides patients into several diagnostic groups for management based on factors such as patient age, disease diagnosis, comorbidities, complications, treatment methods, severity of symptoms and prognosis, and resource consumption. Related studies have found that according to the principle of similar resource consumption in DRGs grouping, however, previous DRGs-related research has been limited to the fairness of hospital human resource allocation, service capabilities, and performance evaluation. There has been no exploration of developing a DRGs nursing human resource allocation model suitable for China's national conditions under the background of DRGs.

[0037] DRGs contains a systematic classification of patients and the corresponding DRGs weight information. The calculation formula for DRGs weight is generally: DRGs weight = average cost of a DRG group / average cost per visit in the province. Through this formula, the weight value of each DRG group can be obtained. The higher the value, the higher the resource consumption of the disease group. Due to differences in medical levels, charging standards and other factors in different regions, DRGs weights may vary between different regions. It has a certain degree of dynamism and can be better applied to a hospital in a certain region.

[0038] DRGs' classification capabilities for patients with mental illness, long-term hospitalization, and rehabilitation are not yet prominent, but are relatively more suitable for departments such as orthopedics and surgery. In this context, the embodiments of the present invention provide an evaluation model training method for an orthopedic nursing human resource allocation model.

[0039] Utilize DRGs and existing factors affecting nursing human resource allocation that are applicable to data indicators captured from hospital information platforms; rely on the hospital's information platform and use machine learning algorithms to build a nursing human resource allocation model; use information and intelligent methods to achieve real-time dynamic optimization of clinical nursing human resource allocation management in wards.

[0040] The nursing difficulty score of the patient needs to form a number of evaluation indicators based on the patient's own physical condition, basic indicators, and corresponding social attribute information. The evaluation indicator in the embodiment of the present invention refers to a certain item and its corresponding information, such as BMI=23, which can be regarded as an evaluation indicator; observe individual DRGs information and nursing information: for example, DRGs weight=6.3, BMI=23, gender: male, age: 34 years old, right calf tibial fracture, etc.

[0041] In the existing evaluation models, there are often a large number of evaluation indicators, which may have a low correlation with the patient's own nursing difficulty, making the evaluation system / model parameters too complicated and even easy to cause model misjudgment. To address this, single-factor analysis and multi-factor analysis are added during the establishment and training of the evaluation system, and the predictive ability of the evaluation model is evaluated to determine whether the evaluation indicators are accurate, so as to clarify the influencing factors of nursing human resource allocation that are suitable for capturing data indicators on the hospital information platform.

[0042] Through literature retrieval, qualitative interviews, and case tracking, we sorted out and summarized the influencing factors / evaluation indicators of human resource allocation and formed a questionnaire. The Delphi method was used to screen the evaluation indicator system of nursing human resource allocation that was suitable for data extraction from the information platform of our center and had good scientificity and clinical accessibility. The principal component analysis method was used to calibrate the evaluation indicator system of nursing human resource allocation with orthopedic characteristics.

[0043] Specifically, an evaluation index system for orthopedic nursing human resource allocation is formed:

[0044] ① According to the Delphi method, 15-20 experts in nursing management, specialist nursing, digital informationization, and machine learning were selected to understand the sensitivity, feasibility, importance, necessity, and scores of the preliminary indicators through a questionnaire. Expert inclusion criteria: bachelor degree or above, senior professional title; engaged in professional practice for ≥15 years; undertake or participate in scientific research projects at the provincial and ministerial levels or above, and publish papers as the first author; can actively participate in and complete the discussion of this study. ② Indicator screening criteria: the mean of the importance and operability of the items is ≥3.5, and the coefficient of variation is ≤20%. Statistical software was used for data analysis, and descriptive analysis was expressed by mean, standard deviation, coefficient of variation (CV), degree of agreement, etc. ③ The group members discussed the selection of indicators based on the experts' written suggestions and preliminarily constructed the evaluation indicators for nursing human resource allocation.

[0045] Correction of the evaluation indicators of orthopedic nursing human resource allocation: Inpatients from different departments, different nursing levels, and different conditions of the Department of Orthopedics were randomly selected, and the constructed nursing human resource allocation indicators were used for investigation. Principal component analysis, LASSO regression, and Ridge regression were used to correct the indicator constituent elements to form a nursing human resource allocation evaluation indicator system with orthopedic characteristics.

[0046] On this basis, we can further construct an orthopedic nursing human resource allocation model based on machine learning: (1) Organize and select clinical observation individuals. (2) Information collection: Use the hospital information platform to capture the corresponding data of the evaluation index system, and use the DRGs weight value of the observation individual as the independent variable; collect the nursing resource allocation demand value as the proxy dependent variable. (3) Data standardization processing: Perform big data cleaning and preprocessing on the collected information, and extract data features for standardization to ensure data integrity and accuracy. (4) Construct an orthopedic nursing human resource allocation prediction model: Take 80% of the full sample as the training set, select the optimal machine learning algorithm, construct a prediction model, and predict the human resource demand value. (5) Verify the orthopedic nursing human resource allocation model: Take 20% of the full sample as the test set, use linear and nonlinear mean square error (MSE) prediction, and optimize and adjust the model.

[0047] Going further: Clinically verify the orthopedic nursing human resource allocation model: (1) Organizational selection of observation individuals. (2) Information collection and data standardization processing. (3) Use correlation analysis to test the practical performance of the model: Analyze the fit between the machine learning model calculation results and the patient DRGs weight value and nursing quality score. (4) Improve and perfect the model through expert interviews, field research, and out-of-sample data testing methods to improve the model's calculation capabilities.

[0048] In some embodiments of the present invention, Figure 1 As shown in Table 1, the difference from the above embodiment is that: the evaluation model training method of the nursing human resource configuration model based on machine learning, several of the evaluation indicators include:

[0049] Patient demographic and sociological indicators: DRGs weight value, DRGs classification information, gender, age, marital status, payment method, source region, BMI value; patient clinical symptoms and signs: consciousness, vital signs, blood sugar, bowel movement frequency, intake and output; patient treatment and disposal: ward, nursing level, critical / serious illness, nursing operation, assisted breathing, CRRT, IABP, ECMO; the vital signs include: body temperature, pulse, respiration, systolic blood pressure, diastolic blood pressure, blood oxygen saturation; the intake and output include: urine volume, amount of vomitus, total intake, total output.

[0050] In these embodiments:

[0051] According to the questionnaire, statistical analysis was performed to obtain a nursing difficulty evaluation index table, see Table 1; a comprehensive evaluation of the patient's nursing difficulty was conducted based on the three first-level indicators of patient demographic and sociological characteristics, patient clinical symptoms and signs, and patient treatment and disposal, as well as their subordinate second- and third-level indicators. These indicators were specifically analyzed to further screen evaluation indicators with a high correlation with the patient's nursing difficulty evaluation.

[0052] Table 1 Evaluation indicators of nursing difficulty for inpatients

[0053]

[0054]

[0055] In some embodiments of the present invention, Figure 1 , Figure 2 As shown in Table 2, the difference from the above embodiment is that: the evaluation model training method of the nursing human resource allocation model based on machine learning, the steps of performing univariate analysis on several evaluation indicators and testing the univariate correlation between each evaluation indicator and the patient nursing difficulty score include:

[0056] Performing univariate analysis on the evaluation indicators one by one, and measuring the evaluation indicators as grouping variables;

[0057] A patient care difficulty score for the observed individual is calculated based on the grouping variable.

[0058] In these embodiments:

[0059] The univariate analysis model was used to test the correlation between each influencing factor and the patient care difficulty score. The influencing factors of patient care difficulty included in this study include: DRGs weight, DRGs classification information, gender, age, region of origin, payment method, marital status, BMI value, consciousness, body temperature, pulse, systolic blood pressure, diastolic blood pressure, respiration, blood oxygen saturation, blood sugar, urine volume, vomitus volume, total intake, total output, bowel movement frequency, nursing operation difficulty score, ward, nursing level, condition, assisted breathing, CRRT, IABP, ECMO, etc. First, the influencing factors included in all evaluation index systems were analyzed one by one through the univariate model, with the aim of examining the separate impact of each influencing factor on the patient care difficulty. The analysis results are shown in Table 2 The univariate analysis results of the patient care difficulty evaluation index. The specific form of the univariate model is as follows:

[0060]

[0061] Among them, Y i is the patient care difficulty score of observed individual (patient) i, X jki is a binary indicator variable for the kth group of the jth factor that affects the difficulty of patient care, that is, if the jth factor of individual (patient) i belongs to group k, then X jki =1, otherwise X jki =0. In this study, all influencing factors are measured as grouping variables. i is a random disturbance term. α1,…,α K is the parameter to be estimated, where α1 is the average nursing difficulty score of patients in the baseline group of influencing factor j. k It represents the difference between the average nursing difficulty score of patients in the kth group of influencing factor j and the average nursing difficulty score of patients in the benchmark group.

[0062] Taking the influencing factor of "payment method" as an example, j = payment method, K = 4 (divided into 4 groups in total), the payment method of "free medical care" belongs to the benchmark group, and the average nursing difficulty score of patients in this group is α1, while the nursing difficulty scores of patients in the "local medical insurance", "non-local medical insurance" and "full fee" groups are α1+α2, α1+α3, and α1+α4, respectively.

[0063] In addition to the univariate analysis, the Wilcoxon rank sum test and Kruskal-Wallis rank sum test were also used to compare the differences in patient care difficulty scores between the groups for each factor. P < 0.05 (bilateral) indicated that the differences in patient care difficulty between the groups were statistically significant.

[0064] Table 2 Results of univariate analysis of patient nursing difficulty evaluation indicators (partial)

[0065]

[0066]

[0067] In some embodiments of the present invention, Figure 1 , Figure 3 As shown in Table 3, the difference from the above embodiment is that: the evaluation model training method of the nursing human resource allocation model based on machine learning, the steps of performing multi-factor analysis on several evaluation indicators and testing the multi-factor correlation between each evaluation indicator and the patient nursing difficulty score include:

[0068] Multifactorial models and machine learning were used to examine the relationship between all the evaluation indicators and the patient care difficulty scores;

[0069] The multi-factor model is configured to: set an optimization objective function to identify important evaluation indicators in the multi-factor model, and compress unimportant evaluation indicators to reduce the number of coefficients to be estimated in the model, so as to obtain a trained multi-factor model.

[0070] In these embodiments:

[0071] The relationship between all J influencing factors and the patient care difficulty score was tested using a multifactor model and the elastic network method (ElasticNet) in machine learning. The elastic network method is a combination of the ridge regression method and the LASSO regression method in classical linear machine learning. It identifies the important factors in the multifactor model by setting an optimization objective function, and compresses the unimportant factors to reduce the number of coefficients to be estimated in the model, thereby achieving the effect of accurate dimensionality reduction. The specific form of the multifactor model is as follows:

[0072]

[0073] Equation (III) is the optimization objective function with the introduction of coefficient penalty term, where α is the weight of the l1 norm penalty term. Compared with the single factor model, the constant term θ1 in equation (II) means the average nursing difficulty score of patients whose influencing factors all belong to the benchmark segment. The meanings of other symbols are the same as those in the single factor model.

[0074] According to the above steps, a multi-factor model was constructed, and the elastic network method was used to solve the parameters. Compared with the existing multi-factor model using the ordinary least squares estimation (OLS) algorithm, the elastic network method used in this study is easier to achieve the consistency of model selection and the asymptotic normality of parameter estimation. The resulting compressed estimate has a lower prediction error and can ensure that the estimation result is almost unaffected by outliers in the sample. A large number of empirical studies have shown that when the data is correlated and unstable, the elastic network method can obtain better estimation results than the ridge regression and LASSO methods. The coefficient estimates of each independent variable in the multi-factor model can be obtained, and then the impact of these factors on the patient care difficulty score can be analyzed. In addition, in order to improve the predictive performance of the evaluation model, this study randomly divided the full sample (1495 case samples) into a training set and a test set at a ratio of 8:2, and used the stepwise regression method for training, in which a subsample with a capacity of 80% of the full sample was used as the training set in the machine learning algorithm. The training set is used to analyze the specific information data of each patient, help determine the parameters of the fitting curve, and build an evaluation model for the difficulty of patient care. Ultimately, the evaluation function of the evaluation model is generated.

[0075] In some embodiments of the present invention, Figure 1 , Figure 4 , which is different from the above embodiment in that: the evaluation model training method of the nursing human resource configuration model based on machine learning, the step of analyzing the prediction performance of the multi-factor model includes:

[0076] Obtaining an ideal patient care difficulty score for the observed individual and a patient care difficulty score for the observed individual calculated by the multifactor model;

[0077] The two were compared and the predicted mean square error of the nursing difficulty score of the observed individuals calculated by the multifactor model was calculated.

[0078] In these embodiments:

[0079] After calculating θ1,θ jk The optimal value of After that, the prediction effect of equation (II) needs to be evaluated. The evaluation method is to use a subsample of 20% of the full sample as the test set in the machine learning algorithm, and calculate the mean square error (MSE) of the prediction result of equation (II). That is, the test set is used to analyze the predictive performance of the patient care difficulty evaluation model constructed by the training set. The calculation method of MSE is as follows: first, use equation (IV) to predict the patient care difficulty of the samples in the test set, that is

[0080]

[0081] Among them, Y i is the nursing expert's score of the difficulty of patient care for the observed individual (patient) i, is the prediction result of the patient care difficulty of the observed individual (patient) i calculated by the evaluation model constructed using the training set. Then, by comparing the expert evaluation score of the patient care difficulty with the prediction result, the prediction mean square error (MSE) of the patient care difficulty score in the test set (testset) is calculated as follows (V):

[0082]

[0083] In some embodiments of the present invention, Figure 1 , Figure 5 As shown in Table 3, the difference from the above embodiment is that: the evaluation model training method of the nursing human resource configuration model based on machine learning, the step of analyzing the prediction performance of the multi-factor model includes:

[0084] Using machine learning to correct the regression coefficients of the multi-factor model,

[0085] Obtaining regression coefficients of several evaluation indicators in the multi-factor model;

[0086] It is determined whether the regression coefficient reaches a significant level. If it reaches a significant level, the evaluation index is retained; if it does not reach a significant level, the evaluation index is eliminated from the multi-factor model.

[0087] In some embodiments of the present invention, Figure 1 As shown, the difference from the above embodiment is that:

[0088] An evaluation model training method for a nursing human resource allocation model based on machine learning is proposed, wherein all evaluation indicators included in the multifactor model have an impact on the difficulty of patient care, and the nursing human resource allocation model evaluation system includes the following indicators: DRGs weight, age, payment method, marital status, BMI, body temperature, pulse, respiration, consciousness state, ward, nursing level, condition, nursing operation, CRRT, and ECOM.

[0089] CRRT, the full name of which is Continuous Renal Replacement Therapy, is a blood purification technology mainly used to treat severe acute renal failure, acute necrotizing pancreatitis, systemic inflammatory response syndrome, multiple organ dysfunction syndrome (MODS) and other critical illnesses.

[0090] In the medical field, ECOM usually refers to the English abbreviation of Extracorporeal Membrane Oxygenation, which is an advanced life support technology used to provide extracorporeal breathing and blood circulation support for patients with severe cardiopulmonary failure.

[0091] Moreover, DRGs weights take into account the patient's disease factors and classify cases into patient groups according to the classification ideas of similar clinical courses and similar resource consumption, which has a good effect of data standardization.

[0092] The results of this study showed that the DRGs weight had a great impact on the nursing workload of orthopedic wards and could serve as an important reference for the evaluation of nursing resources for orthopedic patients.

[0093] In these embodiments:

[0094] The elastic network method in machine learning was used. Specifically, the elastic network method was used and the algorithm was regularized to prevent overfitting in the training process containing a large amount of nursing information and DRGs information. The elastic network method was used to correct the regression coefficient of the multivariate model. It was found that the regression coefficients of gender, systolic blood pressure, diastolic blood pressure, assisted breathing and IABP did not reach the significant level in the statistical analysis, so they were eliminated in the multivariate model analysis. In the multivariate model, each independent variable has a corresponding regression coefficient, which indicates the degree of influence of the independent variable on the dependent variable when other independent variables are kept unchanged.

[0095] Use the elastic network method and improve it with regularization: Elastic network regularization is a regularization method for linear regression models that combines the characteristics of L1 and L2 regularization.

[0096] It can handle multicollinearity issues on datasets with a large number of features and select relevant features.

[0097] Elastic Net Regularization controls the size of the regularization term by weighting the L1 norm and the L.2 norm. The L1 norm can produce sparse solutions (i.e., some coefficients are zero) in some cases, while the 2 norm encourages smoothness of the coefficients. Therefore, Elastic Net Regularization can combine the advantages of L1 (lasso regularization) and L2 (ridge regularization) regularization. It is particularly useful when dealing with data with multicollinearity (high correlation between features).

[0098] The formula for elastic net regularization is:

[0099] [L_ftettatl}=L+\lambdaQ1\sum{i=1}^{n}|w_il+\lambda2\sum{i=1}{n}w_i^2]

[0100] Where: (L) is the original loss function, such as mean squared error. (\lambda_1) is the parameter that controls the strength of L1 regularization. (\lambda_2) is the parameter that controls the strength of 2 regularization. (W_i) are the model weights. (n) is the number of weights.

[0101] Balancing L1 and L2: By adjusting the values ​​of (\lambda_1) and (Vambda_2), a trade-off can be made between L1 and L2 regularization to accommodate different data characteristics. 2. Feature selection: When (\lambda__1) is large, the elastic net tends to perform feature selection, resetting the weights of unimportant features to zero. Stability: When (\lambda_2) is large, the elastic net tends to keep the weights small, increasing the stability of the model.

[0102] In addition, according to the diagnosis results of the variance inflation factor, as shown in Table 3, the variance inflation factors of all evaluation factors are less than 10, that is, the current patient care difficulty evaluation model does not have obvious multicollinearity problems. After stepwise regression analysis, DRGs, age, payment method, marital status, BMI, body temperature, pulse, respiration, consciousness state, ward, nursing level, condition, nursing operation, CRRT, and ECOM were finally included, totaling 15 evaluation indicators.

[0103] Based on the results of machine learning, the evaluation model of orthopedic nursing human resource allocation model is constructed as follows:

[0104] Y (nursing manpower demand value calculated by the model) = a (baseline value) + b1 (weight value 1) × x1 (evaluation index 1) + b2 (weight value 2) × x2 (evaluation index 2) + b3 (weight value 3) × x3 (evaluation index 3) + b4 (weight value 4) × x4 (evaluation index 4) + b5 (weight value 5) × x5 (evaluation index 5) +…………b15 (weight value 15) × x15 (evaluation index 15).

[0105] Table 3 Variance inflation factor results

[0106]

[0107] In some embodiments of the present invention, Figure 1 , Figure 6 , as shown in Table 5, the difference from the above embodiment is that:

[0108] The evaluation model training method of the nursing human resource configuration model based on machine learning, wherein the evaluation index extraction comprises:

[0109] Collect a number of preliminary indicators, evaluate the scores of the preliminary indicators according to the Delphi method, and record the indicators if they meet the preliminary indicator screening criteria; if they do not meet the preliminary indicator screening criteria, eliminate the indicators.

[0110] In these embodiments:

[0111] The evaluation index of nursing difficulty for inpatients first selects several indicators as preliminary indicators, several preliminary indicators and their corresponding recognition rates, eliminates the indicators with low recognition rates, preliminarily screens the preliminary indicators, and provides more accurate evaluation indicators.

[0112] In some embodiments of the present invention, Figure 1 , Figure 7 As shown, the difference from the above embodiment is that:

[0113] The evaluation model training method of the nursing human resource configuration model based on machine learning, wherein the selected observation individuals and nursing information include:

[0114] The observed individuals and their initial care information are collected and selected, the initial care information is cleaned and preprocessed, and data features are extracted for standardization; and the standardized data are divided into standard groups according to grading standards.

[0115] In these embodiments:

[0116] The evaluation indicators will be further refined, such as grading, classifying and segmenting age, gender, BMI value, etc., to achieve the division and use of the data contained in the specific indicators.

[0117] In some embodiments of the present invention, Figure 1 , Figure 8 , Fig. 9 As shown, the difference from the above embodiment is that:

[0118] The difference between the above embodiments is that: an evaluation system for a nursing human resource allocation model based on machine learning, the system includes at least one processor; and a memory storing instructions, which, when executed by at least one processor, implements the steps of the method described in any one of the above embodiments.

[0119] In these embodiments:

[0120] As shown in Table 4, the prediction results of the evaluation model have achieved good prediction performance. The present invention mainly uses information and intelligent methods to optimize the clinical nursing difficulty score of the ward by clarifying the influencing factors of nursing human resource configuration applicable to the data indicators captured by the hospital information platform. By combining traditional evaluation indicators with DRGs indicators, the parameters in the existing human resource evaluation system are streamlined. At the same time, the addition of DRGs indicators improves the correlation between the evaluation system and the existing evaluation system, making the evaluation system more universal.

[0121] Table 4 Evaluation model prediction results

[0122]

[0123] The nursing difficulty is a comprehensive judgment based on clinical experience. The score is set from 1 to 30 points, and the score difference is based on 1 point. The higher the score, the greater the nursing difficulty coefficient.

[0124] The processes and logic flows described in this specification may be performed by one or more programmable computers that execute one or more computer programs to perform functions by operating on input data and generating output. The processes and logic flows may also be performed by special purpose logic circuits, such as FPGAs (field programmable gate arrays) or ASICs (application specific integrated circuits), and the apparatus may also be implemented as special purpose logic circuits.

[0125] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0126] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments described herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0127] Although the present invention has been described in detail above by general description, specific implementation methods and experiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. A method for constructing an evaluation model of a nursing human resource configuration model based on machine learning, characterized in that: include: Select the DRGs information and nursing information of the observed individuals, extract the evaluation indicators, and obtain several evaluation indicators; A univariate analysis was performed on several of the evaluation indicators to test the univariate correlation between each evaluation indicator and the patient nursing difficulty score; Conduct multivariate analysis on several of the evaluation indicators, test the multivariate correlation between each evaluation indicator and the patient care difficulty score, and analyze the predictive performance of the multivariate model; The regression coefficient of the multi-factor model is corrected so that all evaluation indicators included in the multi-factor model have an impact on the patient's nursing difficulty, and a nursing human resource allocation model evaluation model that can calculate the patient's nursing difficulty score is obtained.

2. The method for constructing an evaluation model of a nursing human resource allocation model based on machine learning according to claim 1, characterized in that: All the evaluation indicators included in the multifactor model have an impact on the difficulty of patient care, and the evaluation model of the nursing human resource allocation model includes the following evaluation indicators: DRGs weight, DRGs classification information, age, payment method, marital status, BMI, body temperature, pulse, respiration, consciousness state, ward, nursing level, condition, nursing operation, CRRT, and ECOM.

3. The method for constructing an evaluation model of a nursing human resource allocation model based on machine learning according to claim 1, characterized in that: The steps of performing univariate analysis on several evaluation indicators and testing the univariate correlation between each evaluation indicator and the patient nursing difficulty score include: Performing univariate analysis on the evaluation indicators one by one, and measuring the evaluation indicators as grouping variables; A patient care difficulty score for the observed individual is calculated based on the grouping variables.

4. The method for constructing an evaluation model of a nursing human resource allocation model based on machine learning according to claim 1, characterized in that: The steps of performing multi-factor analysis on the evaluation indicators and testing the multi-factor correlation between each evaluation indicator and the patient nursing difficulty score include: Multifactorial models and machine learning were used to examine the relationship between all the evaluation indicators and the patient care difficulty scores; The multi-factor model is configured to: set an optimization objective function to identify important evaluation indicators in the multi-factor model, and compress unimportant evaluation indicators to reduce the number of coefficients to be estimated in the model, so as to obtain a trained multi-factor model.

5. The method for constructing an evaluation model of a nursing human resource allocation model based on machine learning according to claim 1, characterized in that: The step of analyzing the predictive performance of the multi-factor model includes: Obtaining an ideal patient care difficulty score for the observed individual and a patient care difficulty score for the observed individual calculated by the multifactor model; The two were compared and the predicted mean square error of the nursing difficulty score of the observed individuals calculated by the multifactor model was calculated.

6. The method for constructing an evaluation model of a nursing human resource allocation model based on machine learning according to claim 4, characterized in that: The step of analyzing the predictive performance of the multi-factor model includes: Correcting the regression coefficients of the multi-factor model using machine learning; Obtaining regression coefficients of several evaluation indicators in the multi-factor model; It is determined whether the regression coefficient reaches a significant level. If it reaches a significant level, the evaluation index is retained; if it does not reach a significant level, the evaluation index is eliminated from the multi-factor model.

7. The method for constructing an evaluation model of a nursing human resource allocation model based on machine learning according to claim 1, characterized in that: The extraction of evaluation indicators comprises: Collect a number of preliminary indicators, evaluate the scores of the preliminary indicators, and if the preliminary indicators meet the preliminary indicator screening criteria, record the indicators; if the preliminary indicators do not meet the preliminary indicator screening criteria, eliminate the indicators.

8. The method for constructing an evaluation model of a nursing human resource allocation model based on machine learning according to claim 1, characterized in that: The selection of observation individuals and nursing information includes: Collect and select observation individuals and their initial care information; Performing data cleaning and preprocessing on the initial nursing information, and extracting data for standardization; The standardized data is divided into standard groups according to the grading standards.

9. The method for constructing an evaluation model of a nursing human resource configuration model based on machine learning according to claim 1, characterized in that: Some of the evaluation indicators include: DRGs information includes DRGs weights, DRGs weight grouping information, and patient demographic sociological indicators: gender, age, marital status, payment method, region of origin, and BMI value; patient clinical symptoms and signs: consciousness, vital signs, blood sugar, frequency of bowel movements, and intake and output; patient treatment and disposition: ward, nursing level, critical / serious illness, nursing operations, assisted breathing, CRRT, IABP, and ECMO; the vital signs include: body temperature, pulse, respiration, systolic blood pressure, diastolic blood pressure, and blood oxygen saturation; the intake and output include: urine volume, amount of vomitus, total intake, and total output.

10. An evaluation system for a nursing human resource allocation model based on machine learning, the system comprising at least one processor; and a memory storing instructions, which, when executed by at least one processor, implement the steps of the method described in any one of claims 1-9.

Images