Clinical utility of Todai OncoPanel in the setting of approved comprehensive cancer genomic profiling tests in Japan.

Comprehensive cancer genome profiling (CGP) has been nationally reimbursed in Japan since June 2019. Less than 10% of the patients have been reported to undergo recommended treatment. Todai OncoPanel (TOP) is a dual DNA-RNA panel as well as a paired tumor-normal matched test. Two hundred patients underwent TOP as part of Advanced Medical Care B with approval from the Ministry of Health, Labour and Welfare between September 2018 and December 2019. Tests were carried out in patients with cancers without standard treatment or when patients had already undergone standard treatment. Data from DNA and RNA panels were analyzed in 198 and 191 patients, respectively. The percentage of patients who were given therapeutic or diagnostic recommendations was 61% (120/198). One hundred and four samples (53%) harbored gene alterations that were detected with the DNA panel and had potential treatment implications, and 14 samples (7%) had a high tumor mutational burden. Twenty-two samples (11.1%) harbored 30 fusion transcripts or MET exon 14 skipping that were detected by the RNA panel. Of those 30 transcripts, 6 had treatment implications and 4 had diagnostic implications. Thirteen patients (7%) were found to have pathogenic or likely pathogenic germline variants and genetic counseling was recommended. Overall, 12 patients (6%) received recommended treatment. In summary, patients benefited from both TOP DNA and RNA panels while following the same indication as the approved CGP tests. (UMIN000033647).

StepAdd: A personalized mHealth intervention based on social cognitive theory to increase physical activity among type 2 diabetes patients.

OBJECTIVE: Investigate the preliminary efficacy and feasibility of a personalized mobile health (mHealth) intervention based on social cognitive theory (SCT) to promote physical activity among type 2 diabetes patients via self-monitoring, goal setting, and automatic feedback. METHODS: We conducted a pilot study involving 33 type 2 diabetes patients attending Mitsui Memorial Hospital in Japan using a pre-post evaluation design over 12 weeks. Participants measured daily step count, body weight, and blood pressure at home, with the measurements synchronized with the StepAdd application (app) automatically. Participants used the app to review daily results, update personalized step goals, identify individualized barriers to achieving the step goals, find coping strategies to overcome each barrier, and implement these strategies, thereby building effective coping skills to meet the goals. Pharmacists examined the usage of the app and provided coaching on lifestyle modifications. Ultimately, patients established skills to enhance diabetes self-care by using the app. RESULTS: Daily step count increased dramatically with high statistical significance (p < 0.0001), from a mean of 5436 steps/day to 10,150 steps/day, an 86.7 % increase. HbA1c (p = 0.0001) and BMI (p = 0.0038) also improved. Diabetes self-care in diet, exercise, and foot care as well as self-management behavior, self-regulation, and self-efficacy in achieving daily step goals showed significant improvements. The retention rate of the study was very high, at 97.0 % (n = 32). CONCLUSIONS: A personalized smartphone-based mHealth intervention based on SCT is feasible and effective at promoting physical activity among type 2 diabetes patients. The methodology of the intervention could be readily applied to other patient populations.

EHR2CCAS: A framework for mapping EHR to disease knowledge presenting causal chain of disorders - chronic kidney disease example.

OBJECTIVE: The goal of this work was to capture diseases in patients by comprehending the fine-grained medical conditions and disease progression manifested by transitions in medical conditions. We realize this by introducing our earlier work on a state-of-the-art knowledge presentation, which defines a disease as a causal chain of abnormal states (CCAS). Here, we propose a framework, EHR2CCAS, for constructing a system to map electronic health record (EHR) data to CCAS. MATERIALS AND METHODS: EHR2CCAS is a framework consisting of modules that access heterogeneous EHR to estimate the presence of abnormal states in a CCAS for a patient in a given time window. EHR2CCAS applies expert-driven (rule-based) and data-driven (machine learning) methods to identify abnormal states from structured and unstructured EHR data. It features data-driven approaches for unlocking clinical texts and imputations based on the EHR temporal properties and the causal CCAS structure. This study presents the CCAS of chronic kidney disease as an example. A mapping system between the EHR from the University of Tokyo Hospital and CCAS of chronic kidney disease was constructed and evaluated against expert annotation. RESULTS: The system achieved high prediction performance in identifying abnormal states that had strong agreement among annotators. Our handling of narrative varieties in texts and our imputation of the presence of an abnormal state markedly improved the prediction performance. EHR2CCAS presents patient data describing the temporal presence of abnormal states in CCAS, which is useful in individual disease progression management. Further analysis of the differentiation of transition among abnormal states outputted by EHR2CCAS can contribute to detecting disease subtypes. CONCLUSION: This work represents the first step toward combining disease knowledge and EHR to extract abnormality related to a disease defined as fine-grained abnormal states and transitions among them. This can aid in disease progression management and deep phenotyping.

Automatic Classification of Electronic Nursing Narrative Records Based on Japanese Standard Terminology for Nursing.

In Japan, nursing records are not easily put to secondary use because nursing documentation is not standardized. In recent years, electronic health records have necessitated the creation of Japanese nursing terminology. The purpose of this study was to develop and evaluate an automatic classification system for narrative nursing records using natural language processing technology and machine learning. We collected a week's worth of narrative nursing records from an academic hospital. The authors independently annotated the text data, dividing it into morphemes, the smallest meaningful unit in a language. During preprocessing when creating feature quantities, we used a Japanese tokenizer, MeCab, an open-source morphological parser, and the bag-of-words model. A support vector machine was adopted as a classifier for machine learning. The accuracy was 0.96 and 0.86 on the training set and test set, respectively, and the F value was 0.82. Our findings provide useful information regarding the development of an automatic classification system for Japanese nursing records using nursing terminology and natural language processing techniques.

Highly accurate and explainable detection of specimen mix-up using a machine learning model.

Background Delta check is widely used for detecting specimen mix-ups. Owing to the inadequate specificity and sparseness of the absolute incidence of mix-ups, the positive predictive value (PPV) of delta check is considerably low as it is labor consuming to identify true mix-up errors among a large number of false alerts. To overcome this problem, we developed a new accurate detection model through machine learning. Methods Inspired by delta check, we decided to conduct comparisons with the past examinations and broaden the time range. Fifteen common items were selected from complete blood cell counts and biochemical tests. We considered examinations in which ≥11 among the 15 items were measured simultaneously in our hospital; we created individual partial time-series data of the consecutive examinations with a sliding window size of 4. The last examinations of the partial time-series data were shuffled to generate artificial mix-up cases. After splitting the dataset into development and validation sets, we allowed a gradient-boosting-decision-tree (GBDT) model to learn using the development set to detect whether the last examination results of the partial time-series data were artificial mixed-up results. The model's performance was evaluated on the validation set. Results The area under the receiver operating characteristic curve (ROC AUC) of our model was 0.9983 (bootstrap confidence interval [bsCI]: 0.9983-0.9985). Conclusions The GBDT model was more effective in detecting specimen mix-up. The improved accuracy will enable more facilities to perform more efficient and centralized mix-up detection, leading to improved patient safety.

Willingness of Patients Prescribed Medications for Lifestyle-Related Diseases to Use Personal Health Records: Questionnaire Study.

BACKGROUND: Personal health record (PHR) systems let individuals utilize their own health information to maintain and improve quality of life. Using PHRs is expected to support self-management in patients with lifestyle-related diseases. OBJECTIVE: The aim of this study was to identify predictors of the willingness to use PHRs among patients who are prescribed medications for lifestyle-related diseases. METHODS: We recruited pharmacy patrons, aged 20 years or older, who had received at least one medication indicated for hypertension, dyslipidemia, or diabetes. Participants completed self-administered questionnaires regarding their previous diseases, awareness of health care, experience in using PHRs, willingness to use PHRs, and barriers to using PHRs. Data were analyzed using multivariate logistic regression models. RESULTS: Of the 3708 subjects meeting eligibility criteria, 2307 replies (62.22%) were collected. While only 174 (7.54%) participants had previous PHR experience, 853 (36.97%) expressed willingness to use PHRs. In the multivariate analysis, considering exercise to be important for health management (odds ratio [OR] 1.57, 95% CI 1.12-2.21; P=.009), obtaining medical information from books or magazines (OR 1.23, 95% CI 0.96-1.59; P=.10), and obtaining medical information from the internet (OR 1.45, 95% CI 1.13-1.87; P=.004) were newly identified predictors. These were in addition to known predictors, such as being employed, owning information terminals, and previous PHR experience. CONCLUSIONS: Patients who have an active and positive attitude toward health seem to be more willing to use PHRs. Investigating willingness should contribute to the development of more useful PHRs for self-management among patients prescribed medications for lifestyle-related diseases.

The utilization and challenges of Japan's MID-NET® medical information database network in postmarketing drug safety assessments: A summary of pilot pharmacoepidemiological studies.

PURPOSE: To examine the potential role of Medical Information Database Network (MID-NET® ), a newly established Japanese medical information database network, in postmarketing drug safety assessments through the characterization of its advantages and limitations in five pilot studies. METHODS: The pilot studies were designed to address three major objectives in postmarketing drug safety assessments, ie, the examination of actual drug utilization, the impact of safety-related regulatory actions, and drug-associated risks. The five studies were conducted on the following topics: (a) utilization of codeine-containing products and its relationship with respiratory depression, (b) impact of a Dear Healthcare Professional letter on hypocalcemia incidence associated with denosumab (Ranmark® ) use, (c) risk of acute myocardial infarction associated with antidiabetic drug use, (d) risk of glucose metabolism disorders associated with atypical antipsychotic drug use, and (e) prospective monitoring of abnormal laboratory test results during new drug prescriptions. RESULTS: The pilot studies were successfully conducted and demonstrated the value of MID-NET® in postmarketing drug safety assessments. In particular, the ability to utilize laboratory test results as objective clinical indicators is a major advantage of this database. Potential limitations include a relatively small sample size and a lack of patient-level data linkages among hospitals. CONCLUSIONS: MID-NET® was confirmed to be a valuable database for postmarketing drug safety assessments. The use of laboratory test results to define clinical outcomes may allow more objective and accurate analyses to be conducted. These studies furthered our understanding of the characteristics of MID-NET® , including its advantages and limitations.

Establishment of the MID-NET® medical information database network as a reliable and valuable database for drug safety assessments in Japan.

PURPOSE: To establish a new medical information database network (designated MID-NET® ) to provide real-world data for drug safety assessments in Japan. METHODS: This network was designed and developed by the Ministry of Health, Labour and Welfare and the Pharmaceuticals and Medical Devices Agency in collaboration with 23 hospitals from 10 healthcare organizations across Japan. MID-NET® is a distributed and closed network system that connects all collaborative organizations through a central data center. A wide variety of data are available for analyses, including clinical and administrative information. Several coding standards are used to standardize the data stored in MID-NET® to allow the integration of information originating from different hospitals. A rigorous and consistent quality management system was implemented to ensure that MID-NET® data are of high quality and meet Japanese regulatory standards (good post-marketing study practice and related guidelines). RESULTS: MID-NET® was successfully established as a reliable and valuable medical information database and was officially launched in April 2018. High data quality with almost 100% consistency was confirmed between original data in hospitals and the data stored in MID-NET® . A major advantage is that approximately 260 clinical laboratory test results are available for analysis. CONCLUSIONS: MID-NET® is expected to be a major data source for drug safety assessments in Japan. Experiences and best practices established in MID-NET® may provide a model for the future development of similar database networks.

Evaluating the impact of regulatory action on denosumab-induced hypocalcaemia in Japan.

WHAT IS KNOWN AND OBJECTIVE: Since its introduction in April 2012, denosumab has been administered to approximately 7,300 patients as of August 2012, and 32 cases of serious hypocalcaemia after denosumab administration, including two deaths, have been reported in Japan. A Dear Healthcare Professional Letter of Rapid Safety Communication ('Blue letter') was released to warn about the risks of hypocalcaemia associated with denosumab. The goal of this study therefore was to measure the impact of regulatory action on denosumab-induced hypocalcaemia in Japan by using an electronic medical information database (MID). METHODS: We used two different aggregated data sets based on MIDs (data sets one and two). The patients studied were those who were newly prescribed denosumab or zoledronic acid between April 2012 and September 2014. We assessed four indicators: (a) the proportion of patients with calcium supplementation at the initial denosumab treatment, (b) the proportion of patients who underwent a serum calcium test, (c) the average number of serum calcium tests performed and (d) the prevalence of hypocalcaemia. All indices were aggregated by every 3 months. To evaluate the impact of regulatory action, we used difference in difference (DID) analysis. RESULTS AND DISCUSSION: The proportion of patients with calcium supplementation at the initial denosumab treatment increased year by year in both data sets. The average number of serum calcium tests increased year by year in data set two. There was a significant difference in the prevalence of hypocalcaemia in data set two. This suggests that the estimate of impact of the regulatory action may vary according to the database. In DID analysis, however, significant influences of the regulatory action on combination use with a calcium supplement were detected in both data sets. WHAT IS NEW AND CONCLUSION: There was a significant influence on combination use of denosumab with vitamin D and/or calcium supplement in both data sets. That there was no apparent increase in the prevalence of denosumab-induced hypocalcaemia, suggests that the regulatory action had an impact in the clinical setting studied. Such regulatory actions may play an important role in the promotion of drug safety.

Architecture of the Japan Ischemic Heart Disease Multimodal Prospective Data Acquisition for Precision Treatment (J-IMPACT) System.

The utilization of electronic medical records and multimodal medical data is an ideal approach to build a real-time and precision registry type study with a smaller effort and cost, which may fill a gap between evidence-based medicine and the real-world clinical practice. The Japan Ischemic heart disease Multimodal Prospective data Acquisition for preCision Treatment (J-IMPACT) project aimed to build an clinical data registry system that electronically collects not only medical records, but also multimodal data, including coronary angiography and percutaneous coronary intervention (PCI) report, in standardized data formats for clinical studies.The J-IMPACT system comprises the standardized structured medical information exchange (SS-MIX), coronary angiography and intervention reporting system (CAIRS), and multi-purpose clinical data repository system (MCDRS) interconnected within the institutional network. In order to prove the concept, we acquired multimodal medical data of 6 consecutive cases that underwent PCI through the J-IMPACT system in a single center. Data items regarding patient background, laboratory data, prescriptions, and PCI/cardiac catheterization report were correctly acquired through the J-IMPACT system, and the accuracy of the multimodal data of the 4 categories was 100% in all 6 cases.The application of J-IMPACT system to clinical studies not only fills the gaps between randomized clinical trials and real-world medicine, but may also provide real-time big data that reinforces precision treatment for each patient.

Association between number of institutions with coronary computed tomography angiography and regional mortality ratio of acute myocardial infarction: a nationwide ecological study using a spatial Bayesian model.

BACKGROUND: Coronary computed tomography angiography (CTA) has demonstrated high diagnostic accuracy for detection of coronary artery stenosis, and healthcare providers can detect coronary artery disease in earlier stages before it develops into more serious clinical conditions such as acute myocardial infarction (AMI). We hypothesized that the mortality ratio of AMI in regions with a higher density of coronary CTA is lower than that in regions with a lower density of coronary CTA. METHODS: This ecological and cross-sectional study using secondary data targeted all secondary medical service areas (SMSAs) in Japan (n = 349). We obtained the numbers of cardiologists, institutions with coronary CTA, and institutions with a cardiac catheterization laboratory (CCL) as medical resources, socioeconomic factors, lifestyle factors, exercise habit factors, and AMI mortality data from a Japanese national database. We evaluated the association between the number of these medical resources and the standardized mortality ratio (SMR) of AMI in each SMSA using a hierarchical Bayesian model accounting for spatial autocorrelation (i.e., a conditional autoregressive model). We assumed a Poisson distribution for the observed number of AMI-related deaths and set the expected number of AMI-related deaths as the offset variable. RESULTS: The number of institutions with coronary CTA was negatively and significantly associated with the SMR of AMI (relative risk [RR] 0.900; 95% credible interval [CI] 0.848-0.953), while the SMR in each SMSA was not significantly associated with the number of either cardiologists (RR 0.997; 95% CI 0.988-1.004) or institutions with a CCL (RR 1.026; 95% CI 0.963-1.096). CONCLUSIONS: We observed a significant association between the number of institutions with coronary CTA and the SMR of AMI. Effective allocation of coronary CTA in each region is recommended, and it would be important to clarify the standing position of coronary CTA in regional networking for AMI treatment in the future.

Evaluating the Effectiveness of a Fall Risk Screening Tool Implemented in an Electronic Medical Record System.

We investigated the effect of using a fall risk screening tool in an electronic medical record system by using data for 25 039 patients in 24 general wards of a single institution. The probability of the occurrence of falls decreased after the tool was implemented, but using the tool did not reduce the actual occurrence of falls. This indicates that we must improve not only the assessment of the risk of falls but also the interventions to prevent falls.

A positive family history of hypertension might be associated with an accelerated onset of type 2 diabetes: Results from the National Center Diabetes Database (NCDD-02).

Type 2 diabetes, which is characterized by a combination of decreased insulin secretion and decreased insulin sensitivity, can be delayed or prevented by healthy lifestyle behaviors. Therefore, it is important that the population in general understands their personal risk at an early age to reduce their chances of ever developing the disease. A family history of hypertension is known to be associated with insulin resistance, but the effect of a family history of hypertension on the onset of type 2 diabetes has not well been examined. We performed a retrospective study examining patient age at the time of the diagnosis of type 2 diabetes by analyzing a dataset of 1,299 patients (1,021 men and 278 women) who had been diagnosed as having type 2 diabetes during a health checkup. The mean ± standard deviation of the patient age at the time of the diagnosis of diabetes was 49.1 ± 10.4 years for patients with a family history of hypertension and 51.8 ± 11.4 years for patients without a family history of hypertension (p < 0.001). A multivariate linear regression analysis showed a significant association between a family history of hypertension and a younger age at the time of the diagnosis of type 2 diabetes, independent of a family history of diabetes mellitus and a male sex, suggesting that a positive family history of hypertension might be associated with the accelerated onset of type 2 diabetes.

Establishing a Classification System for High Fall-Risk Among Inpatients Using Support Vector Machines.

We constructed a model using a support vector machine to determine whether an inpatient will suffer a fall on a given day, depending on patient status on the previous day. Using fall report data from our own facility and intensity-of-nursing-care-needs data accumulated through hospital information systems, a dataset comprising approximately 1.2 million patient-days was created. Approximately 50% of the dataset was used as training and testing data. A multistep grid search was conducted using the semicomprehensive combination of three parameters. A discriminant model for the testing data was created for each parameter to identify which parameter had the highest score by calculating the sensitivity and specificity. The score of the model with the highest score had a sensitivity of 64.9% and a specificity of 69.6%. By adopting a method that relies on daily data recorded in the electronic medical record system and accurately predicts unknown data, we were able to overcome issues described in previous studies while simultaneously constructing a discriminant model for patients' fall risk that does not burden nurses and patients with information gathering.

Impact of Doctor Car with Mobile Cloud ECG in reducing door-to- balloon time of Japanese ST-elevation myocardial infarction patients.

Early reperfusion by percutaneous coronary intervention (PCI) is the current standard therapy for ST-elevation myocardial infarction (STEMI). To achieve better prognoses for these patients, reducing the door-to-balloon time is essential. As we reported previously, the Kitasato University Hospital Doctor Car (DC), an ambulance with a physician on board, is equipped with a novel mobile cloud 12-lead ECG system. Between September 2011 and August 2013, there were 260 emergency dispatches of our Doctor Car, of which 55 were for suspected acute myocardial infarction with chest pain and cold sweat. Among these 55 calls, 32 patients received emergent PCI due to STEMI (DC Group). We compared their data with those of 76 STEMI patients who were transported directly to our hospital by ambulance around the same period (Non-DC Group). There were no differences in patient age, gender, underlying diseases, or Killip classification between the two groups. The door-to-balloon time in the DC group was 56.1 ± 13.7 minutes and 74.0 ± 14.1 minutes in the Non-DC Group (P < 0.0001). Maximum levels of CPK were 2899 ± 308 and 2876 ± 269 IU/L (P = 0.703), and those of CK-MB were 292 ± 360 and 295 ± 284 ng/mL (P = 0.423), respectively, in the 2 groups. The Doctor Car system with the Mobile Cloud ECG was useful for reducing the door-to-balloon time.

Perioperative factors affecting the occurrence of acute complex regional pain syndrome following limb bone fracture surgery: data from the Japanese Diagnosis Procedure Combination database.

OBJECTIVE: Complex regional pain syndrome (CRPS) describes a broad spectrum of symptoms that predominantly localize to the extremities. Although limb fracture is one of the most frequently reported triggering events, few large-scale studies have shown the occurrence of and factors associated with CRPS following limb fracture. This study aimed to show the occurrence and identify of those factors. METHODS: Using the Japanese Diagnosis Procedure Combination database, we identified 39 patients diagnosed with CRPS immediately after open reduction and internal fixation (ORIF) for limb fracture from a cohort of 185 378 inpatients treated with ORIF between 1 July and 31 December of each year between 2007 and 2010. Patient and clinical characteristics such as age, gender, fracture site, duration of anaesthesia and use of regional anaesthesia were investigated by logistic regression analyses to examine associations between these factors and the in-hospital occurrence of CRPS after ORIF. RESULTS: The occurrence of CRPS was relatively high in fractures of the distal forearm, but low in fractures of the lower limb and in patients with multiple fractures. Generally females are considered to be at high risk of CRPS; however, we found a comparable number of male and female patients suffering from CRPS after ORIF for limb fracture. In terms of perioperative factors, a longer duration of anaesthesia, but not regional anaesthesia, was significantly associated with a higher incidence of CRPS. CONCLUSION: Although a limited number of CRPS patients were analysed in this study, reduced operative time might help to prevent the development of acute CRPS following limb fracture.

Mortality and morbidity after high-dose methylprednisolone treatment in patients with acute cervical spinal cord injury: a propensity-matched analysis using a nationwide administrative database.

OBJECTIVE: To examine the magnitude of the adverse impact of high-dose methylprednisolone treatment in patients with acute cervical spinal cord injury (SCI). METHODS: We examined the abstracted data from the Japanese Diagnosis Procedure Combination database, and included patients with ICD-10 code S141 who were admitted on an emergency basis between 1 July and 31 December in 2007-2009. The investigation evaluated the patients' sex, age, comorbidities, Japan Coma Scale, hospital volume and the amount of methylprednisolone administered. One-to-one propensity-score matching between high-dose methylprednisolone group (>5000 mg) and control group was performed to compare the rates of in-hospital death and major complications (sepsis; pneumonia; urinary tract infection; gastrointestinal ulcer/bleeding; and pulmonary embolism). RESULTS: We identified 3508 cervical SCI patients (2652 men and 856 women; mean age, 60.8 ± 18.7 years) including 824 (23.5%) patients who received high-dose methylprednisolone. A propensity-matched analysis with 824 pairs of patients showed a significant increase in the occurrence of gastrointestinal ulcer/bleeding (68/812 vs 31/812; p<0.001) in the high-dose methylprednisolone group. Overall, the high-dose methylprednisolone group demonstrated a significantly higher risk of complications (144/812 vs 96/812;OR, 1.66; 95% CI 1.23 to 2.24; p=0.001) than the control group. There was no significant difference in in-hospital mortality between the high-dose methylprednisolone group and the control group (p=0.884). CONCLUSIONS: Patients receiving high-dose methylprednisolone had a significantly increased risk of major complications, in particular, gastrointestinal ulcer/bleeding. However, high-dose methylprednisolone treatment was not associated with any increase in mortality.

Graph Representation Learning-Based Fixed-Length Clinical Feature Vector Generation from Heterogeneous Medical Records.

Transformation of patient data extracted from a database into fixed-length numerical vectors requires expertise in topical medical knowledge as well as data manipulation-thus, manual feature design is labor-intensive. In this study, we propose a machine learning-based method to for this purpose applicable to electronic medical data recorded during hospitalization, which utilizes unsupervised feature extraction based on graph embedding. Unsupervised learning is performed on a heterogeneous graph using Graph2Vec, and the inclusion of clinically useful data in the obtained embedding representation is evaluated by predicting readmission within 30 days of discharge based on it. The embedded representations are observed to improve predictive performance significantly as the information contained in the graph increases, indicating the suitability of the proposed method for feature design corresponding to clinical information.

Immediate postnatal prediction of death or bronchopulmonary dysplasia among very preterm and very low birth weight infants based on gradient boosting decision trees algorithm: A nationwide database study in Japan.

INTRODUCTION: Bronchopulmonary dysplasia (BPD) poses a substantial global health burden. Individualized treatment strategies based on early prediction of the development of BPD can mitigate preterm birth complications; however, previously suggested predictive models lack early postnatal applicability. We aimed to develop predictive models for BPD and mortality based on immediate postnatal clinical data. METHODS: Clinical information on very preterm and very low birth weight infants born between 2008 and 2018 was extracted from a nationwide Japanese database. The gradient boosting decision trees (GBDT) algorithm was adopted to predict BPD and mortality, using predictors within the first 6 h postpartum. We assessed the temporal validity and evaluated model adequacy using Shapley additive explanations (SHAP) values. RESULTS: We developed three predictive models using data from 39,488, 39,096, and 40,291 infants to predict "death or BPD," "death or severe BPD," and "death before discharge," respectively. These well-calibrated models achieved areas under the receiver operating characteristic curve of 0.828 (95% CI: 0.828-0.828), 0.873 (0.873-0.873), and 0.887 (0.887-0.888), respectively, outperforming the multivariable logistic regression models. SHAP value analysis identified predictors of BPD, including gestational age, size at birth, male sex, and persistent pulmonary hypertension. In SHAP value-based case clustering, the "death or BPD" prediction model stratified infants by gestational age and persistent pulmonary hypertension, whereas the other models for "death or severe BPD" and "death before discharge" commonly formed clusters of low mortality, extreme prematurity, low Apgar scores, and persistent pulmonary hypertension of the newborn. CONCLUSIONS: GBDT models for predicting BPD and mortality, designed for use within 6 h postpartum, demonstrated superior prognostic performance. SHAP value-based clustering, a data-driven approach, formed clusters of clinical relevance. These findings suggest the efficacy of a GBDT algorithm for the early postnatal prediction of BPD.

Providing Practical Knowledge and Skills to Handle Real-World Data? Lessons Learned from Med RWD Program.

Both lectures and hands-on education are essential for the development of human resources that can use real-world data (RWD). The University of Tokyo has launched a new hybrid-style RWD educational program entitled "Medical Real World Data Utilization Human Resource Development Project" from FY2019 onwards. We present an overview of the overall picture of the project, including the development process of the educational program and the challenges associated with it.