Stroke Prognostic Scores and Data-Driven Prediction of Clinical Outcomes After Acute Ischemic Stroke.

Background and Purpose- Several stroke prognostic scores have been developed to predict clinical outcomes after stroke. This study aimed to develop and validate novel data-driven predictive models for clinical outcomes by referring to previous prognostic scores in patients with acute ischemic stroke in a real-world setting. Methods- We used retrospective data of 4237 patients with acute ischemic stroke who were hospitalized in a single stroke center in Japan between January 2012 and August 2017. We first validated point-based stroke prognostic scores (preadmission comorbidities, level of consciousness, age, and neurological deficit [PLAN] score, ischemic stroke predictive risk score [IScore], and acute stroke registry and analysis of Lausanne [ASTRAL] score in all patients; Houston intraarterial recanalization therapy [HIAT] score, totaled health risks in vascular events [THRIVE] score, and stroke prognostication using age and National Institutes of Health Stroke Scale-100 [SPAN-100] in patients who received reperfusion therapy) in our cohort. We then developed predictive models using all available data by linear regression or decision tree ensembles (random forest and gradient boosting decision tree) and evaluated their area under the receiver operating characteristic curve for clinical outcomes after repeated random splits. Results- The mean (SD) age of the patients was 74.7 (12.9) years and 58.3% were men. Area under the receiver operating characteristic curves (95% CIs) of prognostic scores in our cohort were 0.92 PLAN score (0.90-0.93), 0.86 for IScore (0.85-0.87), 0.85 for ASTRAL score (0.83-0.86), 0.69 for HIAT score (0.62-0.75), 0.70 for THRIVE score (0.64-0.76), and 0.70 for SPAN-100 (0.63-0.76) for poor functional outcomes, and 0.87 for PLAN score (0.85-0.90), 0.88 for IScore (0.86-0.91), and 0.88 ASTRAL score (0.85-0.91) for in-hospital mortality. Internal validation of data-driven prediction models showed that their area under the receiver operating characteristic curves ranged between 0.88 and 0.94 for poor functional outcomes and between 0.84 and 0.88 for in-hospital mortality. Ensemble models of a decision tree tended to outperform linear regression models in predicting poor functional outcomes but not in predicting in-hospital mortality. Conclusions- Stroke prognostic scores perform well in predicting clinical outcomes after stroke. Data-driven models may be an alternative tool for predicting poststroke clinical outcomes in a real-world setting.

Development of Machine Learning Prediction Models for Self-Extubation After Delirium Using Emergency Department Data.

Delirium is common in the emergency department, and once it develops, there is a risk of self-extubation of drains and tubes, so it is critical to predict delirium before it occurs. Machine learning was used to create two prediction models in this study: one for predicting the occurrence of delirium and one for predicting self-extubation after delirium. Each model showed high discriminative performance, indicating the possibility of selecting high-risk cases. Visualization of predictors using Shapley additive explanation (SHAP), a machine learning interpretability method, showed that the predictors of delirium were different from those of self-extubation after delirium. Data-driven decisions, rather than empirical decisions, on whether or not to use physical restraints or other actions that cause patient suffering will result in improved value in medical care.

Temporal Generalizability of Machine Learning Models for Predicting Postoperative Delirium Using Electronic Health Record Data: Model Development and Validation Study.

BACKGROUND: Although machine learning models demonstrate significant potential in predicting postoperative delirium, the advantages of their implementation in real-world settings remain unclear and require a comparison with conventional models in practical applications. OBJECTIVE: The objective of this study was to validate the temporal generalizability of decision tree ensemble and sparse linear regression models for predicting delirium after surgery compared with that of the traditional logistic regression model. METHODS: The health record data of patients hospitalized at an advanced emergency and critical care medical center in Kumamoto, Japan, were collected electronically. We developed a decision tree ensemble model using extreme gradient boosting (XGBoost) and a sparse linear regression model using least absolute shrinkage and selection operator (LASSO) regression. To evaluate the predictive performance of the model, we used the area under the receiver operating characteristic curve (AUROC) and the Matthews correlation coefficient (MCC) to measure discrimination and the slope and intercept of the regression between predicted and observed probabilities to measure calibration. The Brier score was evaluated as an overall performance metric. We included 11,863 consecutive patients who underwent surgery with general anesthesia between December 2017 and February 2022. The patients were divided into a derivation cohort before the COVID-19 pandemic and a validation cohort during the COVID-19 pandemic. Postoperative delirium was diagnosed according to the confusion assessment method. RESULTS: A total of 6497 patients (68.5, SD 14.4 years, women n=2627, 40.4%) were included in the derivation cohort, and 5366 patients (67.8, SD 14.6 years, women n=2105, 39.2%) were included in the validation cohort. Regarding discrimination, the XGBoost model (AUROC 0.87-0.90 and MCC 0.34-0.44) did not significantly outperform the LASSO model (AUROC 0.86-0.89 and MCC 0.34-0.41). The logistic regression model (AUROC 0.84-0.88, MCC 0.33-0.40, slope 1.01-1.19, intercept -0.16 to 0.06, and Brier score 0.06-0.07), with 8 predictors (age, intensive care unit, neurosurgery, emergency admission, anesthesia time, BMI, blood loss during surgery, and use of an ambulance) achieved good predictive performance. CONCLUSIONS: The XGBoost model did not significantly outperform the LASSO model in predicting postoperative delirium. Furthermore, a parsimonious logistic model with a few important predictors achieved comparable performance to machine learning models in predicting postoperative delirium.

Explanation of machine learning models using shapley additive explanation and application for real data in hospital.

BACKGROUND AND OBJECTIVE: When using machine learning techniques in decision-making processes, the interpretability of the models is important. In the present paper, we adopted the Shapley additive explanation (SHAP), which is based on fair profit allocation among many stakeholders depending on their contribution, for interpreting a gradient-boosting decision tree model using hospital data. METHODS: For better interpretability, we propose two novel techniques as follows: (1) a new metric of feature importance using SHAP and (2) a technique termed feature packing, which packs multiple similar features into one grouped feature to allow an easier understanding of the model without reconstruction of the model. We then compared the explanation results between the SHAP framework and existing methods using cerebral infarction data from our hospital. RESULTS: The interpretation by SHAP was mostly consistent with that by the existing methods. We showed how the A/G ratio works as an important prognostic factor for cerebral infarction using proposed techniques. CONCLUSION: Our techniques are useful for interpreting machine learning models and can uncover the underlying relationships between features and outcome.

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.

A functional learning health system in Japan: Experience with processes and information infrastructure toward continuous health improvement.

Introduction and definition of the term Learning Health System (LHS) appears to have occurred initially around 2007. Prior to this and the introduction of electronic health records (EHR), a predecessor could be found in the Clinical Pathways concept as a standard medical care plan and a tool to improve medical quality. Since 1997, Japan's Saiseikai Kumamoto Hospital (SKH) has been studying and implementing Clinical Pathways. In 2010, they implemented EHR, which facilitated the collection of structured data in common templates that aligned with outcome measurements defined through Japan's Society of Clinical Pathways. For each patient at this hospital, variances from the desired outcomes have been recorded, producing volumes of structured data in formats that could readily be aggregated and analyzed. A visualization tool was introduced to display graphs on the home page of the EHR such that each patient can be compared to similar patients. Knowledge learned from patient care is shared regularly through Clinical Pathways meetings that are supported by all staff within the hospital. The SKH experience over the past two decades is worth exploring further in the context of the development of a fully functional LHS and the attributes/characteristics thereof. In this report, the SKH experience and processes are compared with previously published attributes of a fully functional LHS (ie, characteristics of an LHS that can indicate maturity). Specific examples of the SKH system are detailed with respect to leveraging knowledge gained to change performance that improves patient care as prescribed by learning health cycles. The SKH experience and its information infrastructure and culture exemplify a functional LHS, which is now being expanded to additional hospitals with the hope that it can be scaled and serve as a solid platform for measures aimed at improving medical care, thus establishing broader and more global learning health systems.

Impact of a learning health system on acute care and medical complications after intracerebral hemorrhage.

INTRODUCTION: Patients with stroke often experience pneumonia during the acute stage after stroke onset. Oral care may be effective in reducing the risk of stroke-associated pneumonia (SAP). We aimed to determine the changes in oral care, as well as the incidence of SAP, in patients with intracerebral hemorrhage, following implementation of a learning health system in our hospital. METHODS: We retrospectively analyzed the data of 1716 patients with intracerebral hemorrhage who were hospitalized at a single stroke center in Japan between January 2012 and December 2018. Data were stratified on the basis of three periods of evolving oral care: period A, during which conventional, empirically driven oral care was provided (n = 725); period B, during which standardized oral care was introduced, with SAP prophylaxis based on known risk factors (n = 469); and period C, during which oral care was risk-appropriate based on learning health system data (n = 522). Logistic regression analysis was performed to evaluate associations between each of the three treatment approaches and the risk of SAP. RESULTS: Among the included patients, the mean age was 71.3 ± 13.6 years; 52.6% of patients were men. During the course of each period, the frequency of oral care within 24 hours of admission increased (P < .001), as did the adherence rate to oral care ≥3 times per day (P < .001). After adjustment for confounding factors, a change in the risk of SAP was not observed in period B; however, the risk significantly decreased in period C (odds ratio 0.61; 95% confidence interval 0.43-0.87) compared with period A. These associations were maintained for SAP diagnosed using strict clinical criteria or after exclusion of 174 patients who underwent neurosurgical treatment. CONCLUSIONS: Risk-appropriate care informed by the use of learning health system data could improve care and potentially reduce the risk of SAP in patients with intracerebral hemorrhage in the acute stage.

[Multi-center study for standardisation of perioperative management of radical prostatectomy using clinical pathways].

PURPOSE: Our previous study indicated that there was a great variety in the perioperative management of radical prostatectomy among hospitals. Therefore we conducted a multiinstitutional study in order to standardize the perioperative management using clinical pathways. MATERIALS AND METHODS: The perioperative data on radical prostatectomy were collected between January and December 2004 (Period 1) from eight hospitals. These outcomes were discussed by researchers and the clinical pathway used at each hospital was constructed or revised. The outcomes using these pathways in Period 2 (January 2005 to March 2006) were compared with those in Period 1. RESULTS: Three hundred seventy-eight men in Period 1 and 360 in Period 2 were enrolled in this study. The settings of perioperative management using the new pathways were relatively similar among hospitals. A majority of perioperative management procedures in Period 2 were conducted as described in the new clinical pathways, judging from the median days. The day of initiating the resumption of fluid intake, meals and removing the epidural anesthetic catheter and drain changed from 1.2 +/- 0.7 to 1.3 +/- 1.4 postoperative day, 1.9 +/- 1.2 to 1.8 +/- 1.7, 2.4 +/- 0.7 to 2.5 +/- 0.6, 3.8 +/- 2.5 to 3.8 +/- 2.8, respectively, without a significant difference. The postoperative day of initiating the walking and discontinuing continuous drip infusion and intravenous antibiotics were shortened from 1.9 +/- 0.9 to 1.5 +/- 0.6, 3.7 +/- 2.1 to 3.1 +/- 2.2 and 3.6 +/- 2.0 to 2.5 +/- 2.2, respectively, which did show a significant difference. The day of removing the urethral catheter was changed 9.1 +/- 4.9 to 8.6 +/- 5.4 without significant difference. But the durations of preoperative hospitalization, hospitalization after removing the urethral catheter and postoperative hospitalization were significantly shortened from 3.4 +/- 2.1 to 2.5 +/- 1.0 days, 8.9 +/- 10.1 to 5.6 +/- 3.8, and 17.9 +/- 10.9 to 14.4 +/- 9.1, respectively. CONCLUSIONS: The clinical pathways established or revised at these hospitals after discussing the perioperative management in multiple hospitals were similar, and using such pathways advanced the standardization of peri-operative management after radical prostatectomy.

A Japanese model of disease management.

We started a disease management model, Carna, that includes two programs: one for primary prevention of lifestyle diseases and one for secondary/tertiary prevention of diabetes mellitus. These programs support the family doctor system and education for participants to allow the concept of disease management to take root in Japan. We developed a critical pathway system that can optimize health care of individual participants by matching individual status. This is the core technology of the project. Under the primary prevention program, we can perform the health check-up/ instruction tasks in the 'Tokutei Kenshin', which will start for all Japanese citizens aged 40-74 years in April 2008. In the diabetic program, Carna matches doctors and new patients, prevents patient dropout, supports detection of early-stage complications by distributing questionnaires periodically, and facilitates medical specialists' cooperation with family doctors. Carna promotes periodic medical examinations and quickly provides the result of blood tests to patients. We are conducting a study to assess the medical outcomes and business model. The study will continue until the end of 2007.

[Evolution of common clinical path for transurethral resection of prostate (TURP)].

OBJECTIVES: Improvement of perioperative management for transurethral resection of the prostate (TURP) by revising the common clinical path was investigated in multiple hospitals. METHODS: We discussed perioperative outcomes using the common path in 2002 and revised it for 2003. Changes in perioperative outcomes between 2002 and 2003 and differences among hospitals were compared. RESULTS: There were no statistically significant differences in age, proportion of patients with mildly impaired activity of daily living and/or impaired cognition, general anesthesia, operating duration, resected weight, incidence of intraoperative complications and blood transfusion between 2002 and 2003. Although there were no differences in preoperative hospital stay, re-hospitalization rate and charges for surgery and anesthesia, Foley catheter was removed significantly earlier from postoperative day 2.9 to 2.3 and total medical charge significantly decreased from 43,703 to 39,661 units (1 unit = 10 yen). The incidence of postoperative pyrexia increased from 2.4% to 11.2% in 2003, however, the incidences of epididymitis, postoperative bleeding and postoperative difficulty on micturition remained stable. The average and standard deviation of postoperative hospital stay and total medical charge at each hospital decreased, however, differences among hospitals found in 2002 remained in 2003. CONCLUSION: We found that standardization can be accomplished by discussing perioperative management using a common path in multiple hospitals and revising the path as needed. Common clinical path should be a valid method of advancing standardization in Japan.

[Questionnaire survey of peri-operative management of radical prostatectomy for the educational institutions of Japanese urological association].

OBJECTIVES: The perioperative management of radical prostatectomy in Japan has not been investigated previously. It would be significant to document present management practices. METHODS: A questionnaire survey of 1,213 educational institutions of Japanese urological association was conducted by mail. RESULTS: Questionnaires were returned from 722 (60%) institutions and responses indicated that radical prostatectomy was performed in 657 of these institutions (91%). Admission to hospital was most frequently scheduled on preoperative day 2; walking was most frequently planned to be initiated on postoperative day (POD) 1; beginning of food intake on POD 2; termination of continuous drip infusion and drain removal on POD 3, removal of the Foley catheter on POD 7; intravenous antibiotics for three days; oral antibiotics for 7 days; and discharge from hospital on POD 14. However, the duration of the indwelling drain and Foley catheter, antibiotic administration and hospitalization varied widely. CONCLUSION: There was great variety in the perioperative management of radical prostatectomy among hospitals. We need to obtain consensus on 1) the timing of drain removal, 2) duration of antibiotics, and 3) the timing of catheter removal, in order to uniformly provide medical care of good quality in Japan.

Graph Clustering System for Text-Based Records in a Clinical Pathway.

The progressive digitization of medical records has resulted in the accumulation of large amounts of data. Electronic medical data include structured numerical data and unstructured text data. Although text-based medical record processing has been researched, few studies contribute to medical practice. The analysis of unstructured text data can improve medical processes. Hence, this study presents a clustering approach for detecting typical patient's condition from text-based medical record of clinical pathway. In this approach, the sentences in a cluster are merged to generate a "sentence graph" of the cluster after classified feature word by Louvain method. An analysis of real text-based medical records indicates that sentence graphs can represent the medical treatment and patient's condition in a medical process. This method could help the standardization of text-based medical records and the recognition of feature medical processes for improving medical treatment.

[Questionnaire survey of peri-operative management of transurethral prostate resection in Japan].

OBJECTIVES: We investigated perioperative management for transurethral resection of the prostate (TURP) in Japan. METHODS: The questionnaire survey was conducted in 1,213 educational institutions for urology. RESULTS: The questionnaires were returned from 722 (60%) institutions. Admission to hospital was most frequently scheduled on preoperative day 1; termination of continuous drip infusion, starting meal intake and walking on postoperative day 1; intravenous antibiotics for three days; removal of Foley catheter on postoperative day 4; oral antibiotics for 7 days; and discharge from hospital on postoperative day 7. CONCLUSION: Although hospitalization was 14 days or less at most institutions, several procedures, especially the administration of prophylactic antibiotics, were fairly varied. Discussions from various perspectives might be needed to standardize the perioperative management of TURP in Japan.

Cost-Effectiveness of Maintenance Hemodialysis in Japan.

The cost-effectiveness according to primary disease or dialysis duration has never been analyzed with respect to maintenance hemodialysis (MHD). Study candidates were > 20 years of age and had received hemodialysis for at least 6 months. Hemodialysis patients were prospectively observed for 36 months, and patient utility was assessed based on the Euro-QOL 5-dimensions (EQ-5D), from which the quality adjusted life years (QALYs) were estimated. Medical costs were calculated based on medical service fees. The cost-effectiveness defined as the incremental cost utility ratio (ICUR) was analyzed from a social perspective. A total of 29 patients (mean age; 59.9 ± 13.1 years) undergoing 437 dialysis sessions were analyzed. Utility based upon the EQ-5D score was 0.75 ± 0.21, and the estimated total medical cost for one year of MHD treatment was 4.52 ± 0.88 US$10 000. ICUR was 6.88 ± 4.47 US$10 000/QALY on average, and when comparing ICUR based on the causes of kidney failure, the value for diabetic nephropathy was found to be higher than that for glomerulonephritis (8.17 ± 6.28 vs. 6.82 ± 4.07). ICUR after 36 months observation increased mainly in the patients below 65 years of age (All; P < 0.05, <65; P < 0.01, 65≤; not significant). MHD is a treatment that could improve the socioeconomic state of elderly patients with end-stage kidney disease (ESKD), but the ICUR for diabetic nephropathy was higher than that for glomerulonephritis.