Nursing necessity analysis device

Abstract

To enable estimation of items that may influence a patient's discharge.SOLUTION: A nursing necessity analysis device comprises a contribution item analysis unit 10 that analyzes items whose variations in nursing-necessity values contribute to a patient becoming possible to be discharged, with nursing-necessity values of each item at the time immediately before discharge during a hospitalization period as discharge-possible state item values, and nursing-necessity values of each item during the hospitalization period prior to the time immediately before discharge as discharge-impossible state item values on the basis of patient-specific history information in which transitions of nursing necessity during the hospitalization period are recorded for each item. The device can estimate nursing-necessity items that may be influencing the patient's discharge by analyzing items that may contribute significantly to the patient becoming possible to be discharged on the assumption that the patient becomes possible to be discharged as a result of the nursing-necessity values of each item at times earlier than the time immediately before discharge shifting to the nursing-necessity values at the time immediately before discharge.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a nursing need analysis device, and is particularly suitable for use in a device that analyzes historical information on nursing need recorded during a patient's hospital stay. [Background technology]

[0002] In general, the "Diagnosis Procedure Combination (DPC)" is used as an index for calculating the daily hospitalization cost according to the disease name, severity, treatment details, etc. In addition, the "medical and nursing necessity level" is used as an index for determining the hospitalization cost and measuring the amount of nursing care required to be provided to hospitalized patients. Conventionally, a system is known that predicts the planned discharge date of a patient using a decision tree model based on data on patient information, DPC information, and medical and nursing necessity level information (see, for example, Patent Document 1).

[0003] The invention described in Patent Document 1 was made in response to the problem that while there is a demand for improving the operational efficiency of nursing and bed management operations in hospitals, it is difficult to predict the discharge date of hospitalized patients due to variances that arise depending on the recovery status of hospitalized patients, etc. However, in hospitals that have introduced the system described in Patent Document 1, although they can consider the operation of nursing and bed management operations based on the predicted expected discharge date, they cannot consider operations to improve the expected discharge date based on the output information of the system. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent Publication No. 2021-26358 Summary of the Invention [Problem to be solved by the invention]

[0005] The present invention has been made to solve such problems, and aims to make it possible to estimate factors that may be affecting a patient's discharge from hospital. [Means for solving the problem]

[0006] In order to solve the above-mentioned problems, in this invention, based on patient-specific history information in which changes in nursing need level during the patient's hospitalization are recorded for each item, the nursing need level value for each item immediately before discharge during the hospitalization period is taken as the discharge-possible state item value, and the nursing need level value for each item during the hospitalization period prior to discharge is taken as the discharge-unpossible state item value, and analysis is performed to analyze the items whose changes in nursing need level value contribute to the patient being able to be discharged. [Effects of the Invention]

[0007] According to the present invention configured as described above, it is assumed that discharge becomes possible as a result of the nursing need level value for each item during the hospitalization period prior to immediately before discharge changing to the nursing need level value for each item immediately before discharge, and with regard to the nursing need level for each item, the fluctuation state of which may differ for each patient and item, items that may have the greatest contribution to being able to be discharged are analyzed, making it possible to estimate the nursing need items that may have an impact on the patient's discharge. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a block diagram showing an example of the functional configuration of a nursing need analysis device according to an embodiment of the present invention. [Figure 2] FIG. 1 is a diagram showing nursing need level items A and B. [Figure 3] FIG. 2 is a diagram illustrating an example of history information stored in a history information storage unit; [Figure 4] FIG. 10 is a diagram for explaining dischargeable state item values ​​and discharge-undischargeable state item values. [Figure 5] FIG. 10 is a diagram showing an example of a SHAP value calculated by an index value calculation unit of the present embodiment. [Figure 6] 10 is a diagram showing another example of a SHAP value calculated by the index value calculation unit of the present embodiment. FIG. [Figure 7] 10 is a diagram showing another example of a SHAP value calculated by the index value calculation unit of the present embodiment. FIG. DETAILED DESCRIPTION OF THE INVENTION

[0009] An embodiment of the present invention will be described below with reference to the drawings. Fig. 1 is a block diagram showing an example of the functional configuration of a nursing need analysis device 1 according to this embodiment. As shown in Fig. 1, the nursing need analysis device 1 of this embodiment comprises a contribution item analysis unit 10 as a functional configuration. The contribution item analysis unit 10 comprises, as specific functional configurations, a model generation unit 11 and an index value calculation unit 12. Furthermore, the nursing need analysis device 1 of this embodiment comprises a history information storage unit 20 as a storage medium.

[0010] The nursing need analysis device 1 is configured by a terminal device such as a personal computer. The contribution item analysis unit 10 executes the following process through cooperation between hardware and software provided in the terminal device. For example, the process of the contribution item analysis unit 10 is executed by a program stored in a storage medium such as RAM, ROM, a hard disk, or a semiconductor memory, under the control of a microcomputer configured with a CPU, RAM, ROM, etc.

[0011] The history information storage unit 20 stores history information for each patient in which changes in nursing need level during the patient's hospitalization are recorded for each item. The nursing need level is evaluated by a medical professional every day from the day the patient is admitted to the hospital until the day before the patient is discharged, and information indicating the evaluation results is stored in the history information storage unit 20 as daily history information. Note that the nursing need analysis device 1 may be equipped with a function for inputting information on nursing need levels evaluated by a medical professional, or a separate device different from the nursing need analysis device 1 may be equipped with such a function. If such a function is equipped in a separate device, the nursing need analysis device 1 and the separate device may be connected via a communication network such as a LAN (Local Area Network), so that the nursing need analysis device 1 can acquire history information from the separate device and store it in the history information storage unit 20.

[0012] Nursing need levels include item A, which evaluates the implementation status of medical management and treatment, item B, which evaluates the patient's ADL status (activities of daily living) and level of consciousness, and item C, which evaluates the medical status of surgery, etc. In this embodiment, of the items A to C included in nursing need levels, historical information for items A and B is used for analysis. Figure 2 shows the A and B items of nursing need levels. Item A includes eight items A1 to A8, and item B includes seven items B9 to B15, and a value of 0, 1, or 2 points is recorded for each item depending on the evaluation results by medical professionals.

[0013] Fig. 3 is a diagram schematically illustrating an example of historical information stored in the historical information storage unit 20. As shown in Fig. 3, the historical information storage unit 20 stores, as historical information, the nursing need level for each item evaluated daily for each of multiple patients P1, P2, ... from the day of admission to the day before the day of discharge.

[0014] The contributing item analysis unit 10 analyzes items for which fluctuations in nursing need level values ​​contribute to the patient being able to be discharged, by setting the nursing need level value for each item immediately before discharge during the hospitalization period as the discharge-possible state item value, and setting the nursing need level value for each item during the hospitalization period prior to discharge as the discharge-unpossible state item value, based on the historical information for each patient stored in the historical information storage unit 20. This analysis involves analyzing items for which fluctuations in nursing need level values ​​from the hospitalization period prior to discharge to immediately before discharge may be a factor promoting discharge; in other words, items for which no fluctuations in nursing need level values ​​may be a factor inhibiting discharge.

[0015] FIG. 4 is a diagram illustrating the discharge possibility status item value and the discharge non-possible status item value. As shown in FIG. 4, for each of multiple patients P1, P2, ..., the nursing need level of each item A1 to B15, evaluated daily from the day of admission to the day before the discharge date, is stored as historical information in the history information storage unit 20. Of these, the value of the nursing need level of each item A1 to B15 on the day before discharge (DH) is the discharge possibility status item value. Furthermore, the value of the nursing need level of each item A1 to B15 during the period SH from the day of admission to two days before discharge is the discharge non-possible status item value. The discharge possibility status item value and the discharge non-possible status item value are set in this manner on the assumption that the value of the nursing need level of each item A1 to B15 during the period SH on the day before discharge changes to the value of the nursing need level of each item A1 to B15 on the day before discharge (DH), resulting in discharge becoming possible.

[0016] Here, the discharge-unavailable status item value is the maximum value for each item among the nursing need level values ​​for each item A1 to B15 in the period SH before the day before. That is, the maximum value among the nursing need level values ​​for item A1 recorded by date and time during the period SH before the day before is the discharge-unavailable status item value for item A1. Similarly, the maximum value among the nursing need level values ​​for item A2 recorded by date and time during the period SH before the day before is the discharge-unavailable status item value for item A2. The same applies to the other items A3 to B15.

[0017] In addition, the value of the nursing need level for each item A1 to B15 on the day the scheduled discharge date is determined may be used as the discharge-possible status item value, and the maximum value of the nursing need level for each item A1 to B15 in the period prior to the day the scheduled discharge date is determined may be used as the discharge-unpossible status item value.

[0018] The contributing item analysis unit 10 executes the processing of the model generation unit 11 and the index value calculation unit 12 in relation to the analysis of items that may be factors promoting (or inhibiting) discharge. The model generation unit 11 performs machine learning using the discharge feasibility status item values ​​of multiple patients recorded in the patient-specific historical information as positive or negative examples, and the discharge infeasibility status item values ​​of multiple patients as negative or positive examples, and generates a determination model that outputs information representing the degree of similarity between the nursing need level of each item input as a determination target and the positive examples. That is, the model generation unit 11 generates a determination model that uses the values ​​of each item A1 to B15 as explanatory variables and a probability value representing the degree of similarity as a response variable.

[0019] The form of the decision model generated by the model generation unit 11 may be any form such as a regression model, a tree model, a neural network model, a Bayesian model, a clustering model, etc. For example, a learning model based on a decision tree, specifically XGBoost (eXtreme Gradient Boosting), etc. may be used.

[0020] When using the nursing need level history information as learning data, the nursing need level values ​​may be converted into dummy variables, and the dummy variables may be used as explanatory variables. For example, the dummy variable value for an item value with a nursing need level of 0 points is set to "0," and the dummy variable value for an item value with a nursing need level of 1 or 2 points is set to "1."

[0021] The index value calculation unit 12 calculates an index value indicating which of the nursing need level items A1 to B15 affects the similarity determination in the determination model generated by the model generation unit 11. As an example, the index value calculation unit 12 calculates the well-known SHAP (SHapley Additive exPlanations) value. The SHAP value is a well-known index value that indicates how each explanatory variable affects the prediction of the value predicted by the learning model.

[0022] 5 is a diagram showing an example of a SHAP value calculated by the index value calculation unit 12. Here, for some items of nursing need level, the SHAP values, which indicate the degree of influence (degree of contribution) that fluctuations in item values ​​have on the calculation of the degree of approximation (objective variable), are shown in order from the item with the greatest degree of contribution. In FIG. 5, gray represents the SHAP value when nursing need level = 0 points (dummy variable = 0), and black represents the SHAP value when nursing need level ≥ 1 point (dummy variable = 1).

[0023] From the positive and negative SHAP values ​​shown in Figure 5, it can be seen that for the three items of "Transferring," "Putting on and taking off clothes," and "Turning over," a nursing need level of 0 contributes positively (the number of patients whose nursing need level moves from 1 or more to 0 from the period SH before the previous day to DH the day before discharge is large), and a nursing need level of ≥ 1 contributes negatively (the number of patients who are discharged with a nursing need level of ≥ 1 is small).From the SHAP values ​​of the other four items, it can be seen that a nursing need level of ≥ 1 contributes positively (a relatively large number of patients are able to be discharged even with a nursing need level of ≥ 1).

[0024] Fig. 6 is a diagram showing another example of the SHAP value calculated by the index value calculation unit 12. Here, an example is shown in which the absolute value average of the SHAP value shown in Fig. 5 is calculated for each item. From the absolute value average shown in Fig. 6, it can be seen that the two items "transferring" and "putting on and taking off clothes" are particularly important items in terms of enabling a patient to be discharged from the hospital.

[0025] The above analysis may be performed for each patient's admission condition. Examples of the patient's admission condition include planned admission, planned readmission (chemotherapy), emergency medical admission, and unplanned admission other than emergency medical admission. When performing the analysis for each patient's admission condition, the model generation unit 11 classifies the patient's history information for each patient's admission condition, and performs machine learning for each admission condition using the classified history information, thereby generating a determination model for each admission condition. The index value calculation unit 12 calculates an index value for each admission condition for the determination model for each admission condition generated by the model generation unit 11.

[0026] FIG. 7 is a diagram showing an example of SHAP values ​​(average absolute values) calculated by the index value calculation unit 12 for each admission condition. From the analysis results shown in FIG. 7, it can be seen that the SHAP value varies depending on the admission condition. For example, in the "specialized treatment / procedure" item, the SHAP value increases in the case of planned admission and planned readmission (the importance of the item increases), and decreases in the case of emergency admission and unplanned admission other than emergency medical admission (the importance of the item decreases). It can also be seen that in the case of planned readmission, item B of nursing need level has almost no importance.

[0027] The four admission conditions shown here are merely examples, and the present invention is not limited to these. For example, the admission conditions may include whether or not an emergency transport is required, whether or not surgery is performed, etc.

[0028] As explained in detail above, in this embodiment, based on patient-specific history information in which changes in nursing need level during the patient's hospitalization are recorded for each item, the value of the nursing need level for each item immediately before discharge during the hospitalization period is taken as the discharge-possible state item value, and the value of the nursing need level for each item during the hospitalization period prior to discharge is taken as the discharge-unpossible state item value, and the items whose changes in the nursing need level value contribute to the patient being able to be discharged are analyzed by calculating the SHAP value.

[0029] According to this embodiment, the nursing need levels of each of the items A1-B15, which may fluctuate differently for each patient and item, are analyzed to determine which items may contribute more to the patient's discharge. This makes it possible to estimate which nursing need items may be affecting the patient's discharge. This makes it possible to clarify how each of the nursing need levels A1-B15 contributes to the patient's discharge, i.e., the importance of each item. Based on the results of this analysis, the hospital can consider operations to improve the patient's expected discharge date, such as prioritizing treatments related to specific nursing needs.

[0030] 5 to 7 show the results of an analysis in which the three nursing need levels (0 / 1 / 2) were replaced with a binary dummy variable (0 / 1), but the analysis can also be performed using the three nursing need levels without replacing them with dummy variables. In this case, it is possible to analyze the degree to which items that caused the nursing need level to decrease from 2 points during the period SH to 1 point on the day before discharge DH, or items that decreased from 1 point to 0 point, contributed to the patient being able to be discharged.

[0031] In the above embodiment, an example was described in which a determination model was generated by machine learning using historical information on nursing need levels, and items that may be promoting (or inhibiting) discharge were analyzed by calculating a SHAP value, but the present invention is not limited to this. That is, the present invention is not limited to analysis methods using machine learning, and other analysis methods may also be used.

[0032] For example, the contribution item analysis unit 10 may calculate, for each of the nursing need level items A1 to B15, the number of patients whose nursing need level values ​​on the day before discharge (DH) were reduced compared to their nursing need level values ​​during the period SH up to the day before discharge, as information indicating the degree to which each of the items A1 to B15 contributed to the patient being able to be discharged. Here, the items may be output in descending order of the number of patients whose nursing need level was reduced. Alternatively, the items for which the number of patients whose nursing need level was reduced equals or exceeds a threshold may be output.

[0033] As another analysis method, the contribution item analysis unit 10 may calculate, for each of the nursing need level items A1 to B15, the average value of the difference between the nursing need level value on the day before discharge (DH) and the nursing need level value during the period SH prior to the day before discharge for multiple patients, as information indicating the degree to which each of the items A1 to B15 contributes to the patient's discharge. Here, the items may be output in descending order of average difference. Alternatively, the items whose average difference is equal to or greater than a threshold may be output.

[0034] Furthermore, the above-described embodiments are merely examples of specific embodiments for carrying out the present invention, and the technical scope of the present invention should not be construed as being limited thereby. In other words, the present invention can be carried out in various forms without departing from its spirit or main characteristics. [Explanation of symbols]

[0035] 1 Nursing necessity analyzer 10 Contribution Item Analysis Department 11 Model Generation Unit 12 Index value calculation section 20 History information storage unit

Claims

1. a contributing item analysis unit that analyzes items to which fluctuations in the nursing need values ​​contribute to the patient being able to be discharged, based on patient-specific history information in which changes in nursing need levels during the patient's hospitalization are recorded for each item, and that sets the value of the nursing need level for each item immediately before discharge during the hospitalization period as a discharge-possible state item value, and the value of the nursing need level for each item during the hospitalization period prior to the discharge as a discharge-unpossible state item value,

2. The above contribution item analysis section a model generation unit that performs machine learning using the dischargeable status item values ​​of the multiple patients recorded in the patient-specific history information as positive or negative examples, and the discharge-unavailable status item values ​​of the multiple patients as negative or positive examples, and generates a judgment model that outputs information representing the degree of similarity between the nursing need level of each item input as a judgment target and the positive examples; an index value calculation unit that calculates an index value indicating how each item of the nursing need level affects the determination of the degree of similarity in the determination model generated by the model generation unit; The nursing need analysis device according to claim 1 .

3. the model generation unit classifies the patient-specific history information by the patient's hospitalization condition, and performs machine learning for each of the hospitalization conditions using the classified history information to generate a determination model for each of the hospitalization conditions; The index value calculation unit calculates the index value for each of the hospitalization conditions for the determination model for each of the hospitalization conditions generated by the model generation unit. The nursing need analysis device according to claim 2 .

4. The nursing need analysis device according to any one of claims 1 to 3, characterized in that the discharge-ineligible state item value is the maximum value for each item among the nursing need values ​​for each item during the hospitalization period prior to the discharge.

5. 2. The nursing need analysis device according to claim 1, wherein the contributing item analysis unit calculates, for each nursing need item, the number of patients whose nursing need value immediately before discharge during the hospitalization period has decreased compared to the nursing need value during the hospitalization period immediately before discharge, as information indicating the degree to which each item contributes to the patient being able to be discharged.

6. 2. The nursing need analysis device according to claim 1, wherein the contribution item analysis unit calculates, for each nursing need item, an average value for a plurality of patients of the difference between the value of the nursing need level immediately before discharge during the hospitalization period and the value of the nursing need level during a hospitalization period prior to immediately before discharge, as information indicating the degree to which each item contributes to the patient being able to be discharged.

Images