We propose prediction model of HbA1c with interpolated missing value based on medical records in order to enable state-space model to follow longitudinal HbA1c variation. The value and the variation of HbA1c are defined as target parameter of the model. The likelihood function is defined by the observed HbA1c value and clustering result of deep learning with text data in the missing value case. The prediction experiments for longitudinal HbA1c data is executed for Kochi Medical School hospital data. The experiment results show that the RMSE of HbA1c after 1,200 days from first test is 0.30, which the result is reduced by about half compared to the method without interpolation process. The proposed method with combining multiple data source is useful for long term prediction of real medical data.

　Development of evidence-based clinical practice guidelines in Japan became more active from around the year 2000. The objective of the present study was to clarify the present state of how clinical practice guidelines are published on the websites of academic societies of clinical medicine in Japan. In total, 355 academic societies were studied to determine whether they had created clinical practice guidelines. In the analysis, the academic societies were divided into two groups based on whether any clinical practice guidelines were listed in the Medical Information Network Distribution Service (Minds). Seven items were compared between the two groups to determine the present state of how clinical practice guidelines are published. Of the 355 academic societies, 108 (30.4％) had developed clinical practice guidelines. Academic societies whose guidelines were listed in Minds had many members, and a page link to the clinical practice guideline was shown on the index page of their websites (difference in proportions=27.3％, P=0.001). Many academic societies only published PDF files of their clinical practice guidelines on their websites. It is hoped that academic societies will publish their clinical practice guidelines on their websites, while taking into account utilization of the guidelines and user convenience.

　In recent dispensing tasks, efficiency and safety are improved by using various support systems. However, there are few reports on the development of a support system that manages all operations collectively in hospital formulation preparation. Therefore, we have developed the “task support system of hospital formulation” for the purpose of collectively managing the hospital formulation preparation tasks and evaluated the usefulness. The database was configured with raw materials, intermediate products, pharmaceutical products, and equipment masters. The system has a function that enables barcode entry/exit and lot management. The master of the preparation procedure was registered, and the standardization of the preparation was attempted by introducing the weight inspection and the imaging function. After preparation, a label printed with information on hospital formulations and a preparation record sheet are automatically issued to enable inventory control. As an evaluation of the system, the time required for the pharmacist to prepare the hospital formulation when the system was used and not used was measured. The preparation time required when using the system was significantly reduced for all item compared to when not using. It was considered that as a result of the introduction of this system, safety was secured and task operations were improved by standardizing the preparation procedure.

　Since 2017, we have been developing an Integrated Cancer Treatment Database System (Cancer DB) which automatically collects confirmed and finalized treatment data from multiple information systems. Surgery data consists of several data items such as patient background, preoperative state, intraoperative remarks, postoperative diagnosis and pathological information which are organized mainly for physicians. We have implemented an NCD output conversion function for thoracic surgical oncology region, which aims to reduce NCD registration burden. To predict NCD surgical procedure code, we defined NCD surgical procedure code classification to apply XGBoost for primary lung cancer (344 cases) and metastatic lung tumor (139 cases), which came with 98％ accuracy. Adding lymph node dissection and lesion count features, NCD surgical procedure code is predictable with high accuracy.

　In recent years, barcodes have come to be printed on the package units for dispensing of all medicines as part of drug safety measures. Of these, the GS1 data bar printed on PTP packaging is in a special printing environment. We examined the factors that affect the readability of barcodes. As a result, “framed”, “printing in sheet center”, and “pitch printing” were significant reading factors as the printing environmental factors. As a bar code quality characteristic factor, “ECmin” and “Decode” were significant readable factors. In the ISO grade classification of the bar code quality, the readability threshold was confirmed in the range of higher grades such as grades A and B. The readability of GS1 data bars printed under special environments such as PTP packaging may be difficult to predict from existing indicators and quality grade classifications.