Machine learning for classification of postoperative patient status using standardized medical data.

BACKGROUND AND OBJECTIVE: Real-world evidence is defined as clinical evidence regarding the use and potential benefits or risks of a medical product derived from real-world data analyses. Standardization and structuring of data are necessary to analyze medical real-world data collected from different medical institutions. An electronic message and repository have been developed to link electronic medical records in this research project, which has simplified the data integration. Therefore, this paper proposes an analysis method and learning health systems to determine the priority of clinical intervention by clustering and visualizing time-series and prioritizing patient outcomes and status during hospitalization. METHODS: Common data items for reimbursement (Diagnosis Procedure Combination [DPC]) and clinical pathway data were examined in this project at each participating institution that runs the verification test. Long-term hospitalization data were analyzed using the data stored in the cloud platform of the institutions' repositories using multiple machine learning methods for classification, visualization, and interpretation. RESULTS: The ePath platform contributed to integrate the standardized data from multiple institutions. The distribution of DPC items or variances could be confirmed by clustering, temporal tendency through the directed graph, and extracting variables that contributed to the prediction and evaluation of SHapley Additive Explanation effects. Constipation was determined to be the risk factor most strongly related to long-term hospitalization. Drainage management was identified as a factor that can improve long-term hospitalization. These analyses effectively extracted patient status to provide feedback to the learning health system. CONCLUSIONS: We successfully generated evidence of medical processes by gathering patient status, medical purposes, and patient outcomes with high data quality from multiple institutions, which were difficult with conventional electronic medical records. Regarding the significant analysis results, the learning health system will be used on this project to provide feedback to each institution, operate it for a certain period, and analyze and re-evaluate it.

Kidney Outcomes Associated With SGLT2 Inhibitors Versus Other Glucose-Lowering Drugs in Real-world Clinical Practice: The Japan Chronic Kidney Disease Database.

OBJECTIVE: Randomized controlled trials have shown kidney-protective effects of sodium-glucose cotransporter 2 (SGLT2) inhibitors, and clinical practice databases have suggested that these effects translate to clinical practice. However, long-term efficacy, as well as whether the presence or absence of proteinuria and the rate of estimated glomerular filtration rates (eGFR) decline prior to SGLT2 inhibitor initiation modify treatment efficacy among type 2 diabetes mellitus (T2DM) and chronic kidney disease (CKD) patients, is unknown. RESEARCH DESIGN AND METHODS: Using the Japan Chronic Kidney Disease Database (J-CKD-DB), a nationwide multicenter CKD registry, we developed propensity scores for SGLT2 inhibitor initiation, with 1:1 matching with patients who were initiated on other glucose-lowering drugs. The primary outcome included rate of eGFR decline, and the secondary outcomes included a composite outcome of 50% eGFR decline or end-stage kidney disease. RESULTS: At baseline, mean age at initiation of the SGLT2 inhibitor (n = 1,033) or other glucose-lowering drug (n = 1,033) was 64.4 years, mean eGFR was 68.1 mL/min per 1.73 m2, and proteinuria was apparent in 578 (28.0%) of included patients. During follow-up, SGLT2 inhibitor initiation was associated with reduced eGFR decline (difference in slope for SGLT2 inhibitors vs. other drugs 0.75 mL/min/1.73 m2 per year [0.51 to 1.00]). During a mean follow-up of 24 months, 103 composite kidney outcomes occurred: 30 (14 events per 1,000 patient-years) among the SGLT2 inhibitors group and 73 (36 events per 1,000 patient-years) among the other drugs group (hazard ratio 0.40, 95% CI 0.26-0.61). The benefit provided by SGLT2 inhibitors was consistent irrespective of proteinuria and rate of eGFR decline before initiation of SGLT2 inhibitors (P heterogeneity ≥ 0.35). CONCLUSIONS: The benefits of SGLT2 inhibitors on kidney function as observed in clinical trials translate to patients treated in clinical practice with no evidence that the effects are modified by the underlying rate of kidney function decline or the presence of proteinuria.

Big data and real-world data-based medicine in the management of hypertension.

Big data has been a hot topic in medical and healthcare research. Big data in healthcare is considered to comprise massive amounts of information from various sources, including electronic health records (EHRs), administrative or claims data, and data from self-monitoring devices. Biomedical research has also generated a significant portion of big data relevant to healthcare. Other large datasets arise from cohorts that are recruited and followed on the basis of specific questions, although such research questions may later be expanded to enable other investigations. While the availability of big data offers many possibilities for an improved understanding of disease and treatment, the need for careful and productive use of statistical concepts should be kept in mind. Patient data routinely collected via electronic means are called real-world data (RWD) and are becoming common in healthcare research. RWD and big data are not synonymous with each other, but the two terms seem to be used without distinction with respect to observational studies. In this article, we review hypertension-related papers that use big data or RWD. There are many other sources of big data or RWD that are not covered here, each of which may pose special challenges and opportunities. While randomized clinical trials (RCTs) are considered to be the criterion standard for generating clinical evidence, the use of real-world evidence (RWE) to evaluate the efficacy and safety of medical interventions is gaining interest. On-going efforts to make use of RWD to generate RWE for regulatory decisions, as well as the challenges confronted, including reliability (quality) and relevance (fitness for purpose) of data, will also be addressed.

Prevalences of hyperuricemia and electrolyte abnormalities in patients with chronic kidney disease in Japan: A nationwide, cross-sectional cohort study using data from the Japan Chronic Kidney Disease Database (J-CKD-DB).

BACKGROUND: The Japan Chronic Kidney Disease Database (J-CKD-DB) is a nationwide clinical database of patients with chronic kidney disease (CKD) based on electronic health records. The objective of this study was to assess the prevalences of hyperuricemia and electrolyte abnormalities in Japanese patients with CKD. METHODS: In total, 35,508 adult outpatients with estimated glomerular filtration rates of 5-60 ml/min/1.73 m2 in seven university hospitals were included this analysis. The proportions of patients with CKD stages G3b, G4, and G5 were 23.5%, 7.6%, and 3.1%, respectively. RESULTS: Logistic regression analysis showed that prevalence of hyperuricemia was associated with CKD stages G3b (adjusted odds ratio [95% confidence interval]: 2.12 [1.90-2.37]), G4 (4.57 [3.92-5.32]), and G5 (2.25 [1.80-2.80]). The respective prevalences of hyponatremia, hypercalcemia, hyperphosphatemia, and narrower difference between serum sodium and chloride concentrations were elevated in patients with CKD stages G3b, G4, and G5, compared with those prevalences in patients with CKD stage G3a. The prevalences of hyperkalemia were 8.3% and 11.6% in patients with CKD stages G4 and G5, respectively. In patients with CKD stage G5, the proportions of patients with optimal ranges of serum uric acid, potassium, corrected calcium, and phosphate were 49.6%, 73.5%, 81.9%, and 56.1%, respectively. CONCLUSIONS: We determined the prevalences of hyperuricemia and electrolyte abnormalities in Japanese patients with CKD using data from a nationwide cohort study.

Prevalence of anemia in patients with chronic kidney disease in Japan: A nationwide, cross-sectional cohort study using data from the Japan Chronic Kidney Disease Database (J-CKD-DB).

BACKGROUND: The Japan Chronic Kidney Disease Database (J-CKD-DB) is a nationwide clinical database of patients with chronic kidney disease (CKD) based on electronic health records. The objective of this study was to assess the prevalence of anemia and the utilization rate of erythropoiesis-stimulating agents (ESAs) in Japanese patients with CKD. METHODS: In total, 31,082 adult outpatients with estimated glomerular filtration rates of 5-60 ml/min/1.73 m2 in seven university hospitals were included this analysis. The proportions of patients with CKD stages G3b, G4, and G5 were 23.5%, 7.6%, and 3.1%, respectively. RESULTS: The mean (standard deviation) hemoglobin level of male patients was 13.6 (1.9) g/dl, which was significantly higher than the mean hemoglobin level of female patients (12.4 (1.6) g/dl). The mean (standard deviation) hemoglobin levels were 11.4 (2.1) g/dl in patients with CKD stage G4 and 11.2 (1.8) g/dl in patients with CKD stage G5. The prevalences of anemia were 40.1% in patients with CKD stage G4 and 60.3% in patients with CKD stage G5. Logistic regression analysis showed that diagnoses of CKD stage G3b (adjusted odds ratio [95% confidence interval]: 2.32 [2.09-2.58]), G4 (5.50 [4.80-6.31]), and G5 (9.75 [8.13-11.7]) were associated with increased prevalence of anemia. The utilization rates of ESAs were 7.9% in patients with CKD stage G4 and 22.4% in patients with CKD stage G5. CONCLUSIONS: We determined the prevalence of anemia and utilization rate of ESAs in Japanese patients with CKD using data from a nationwide cohort study.

J-CKD-DB: a nationwide multicentre electronic health record-based chronic kidney disease database in Japan.

The Japan Chronic Kidney Disease (CKD) Database (J-CKD-DB) is a large-scale, nation-wide registry based on electronic health record (EHR) data from participating university hospitals. Using a standardized exchangeable information storage, the J-CKD-DB succeeded to efficiently collect clinical data of CKD patients across hospitals despite their different EHR systems. CKD was defined as dipstick proteinuria ≥1+ and/or estimated glomerular filtration rate <60 mL/min/1.73 m2 base on both out- and inpatient laboratory data. As an initial analysis, we analyzed 39,121 CKD outpatients (median age was 71 years, 54.7% were men, median eGFR was 51.3 mL/min/1.73 m2) and observed that the number of patients with a CKD stage G1, G2, G3a, G3b, G4 and G5 were 1,001 (2.6%), 2,612 (6.7%), 23,333 (59.6%), 8,357 (21.4%), 2,710 (6.9%) and 1,108 (2.8%), respectively. According to the KDIGO risk classification, there were 30.1% and 25.5% of male and female patients with CKD at very high-risk, respectively. As the information from every clinical encounter from those participating hospitals will be continuously updated with an anonymized patient ID, the J-CKD-DB will be a dynamic registry of Japanese CKD patients by expanding and linking with other existing databases and a platform for a number of cross-sectional and prospective analyses to answer important clinical questions in CKD care.

EBM metadata based on Dublin Core better presenting validity of clinical trials.

To help clinicians find better evidence, a metadata schema for Evidence Based Medicine (EBM) is developed. Dublin Core metadata standard (DC) was adopted to help build a metadata schema. An experimental system was developed to test the validity of the metadata and full text papers of clinical therapy on stomach ulcer extracted using PubMed. An EBM metadata schema was developed. Citations were created from original papers using the metadata schema. Three clinicians evaluated papers by utilizing metadata and full texts respectively. Agreement of evaluation was analyzed, and the result on weighted kappa was 0.55 (95% CI, 0.42-0.67). It reveals that there is moderate agreement between evaluation of metadata citations and full texts. It is possible to use the metadata to select papers before reading the full texts. A further study should be made to prove the applicability of the metadata in the real world setting.

Conceptual model of health information ethics as a basis for computer-based instructions for electronic patient record systems.

A computer-based learning system called Electronic Patient Record (EPR) Laboratory has been developed for students to acquire knowledge and practical skills of EPR systems. The Laboratory is basically for self-learning. Among the subjects dealt with in the system is health information ethics. We consider this to be of the utmost importance for personnel involved in patient information handling. The variety of material on the subject has led to a problem in dealing with it in a methodical manner. In this paper, we present a conceptual model of health information ethics developed using UML to represent the semantics and the knowledge of the domain. Based on the model, we could represent the scope of health information ethics, give structure to the learning materials, and build a control mechanism for a test, fail and review cycle. We consider that the approach is applicable to other domains.