Predicting pharmacotherapeutic outcomes for type 2 diabetes: An evaluation of three approaches to leveraging electronic health record data from multiple sources.

Electronic health record (EHR) data are increasingly used to develop prediction models to support clinical care, including the care of patients with common chronic conditions. A key challenge for individual healthcare systems in developing such models is that they may not be able to achieve the desired degree of robustness using only their own data. A potential solution-combining data from multiple sources-faces barriers such as the need for data normalization and concerns about sharing patient information across institutions. To address these challenges, we evaluated three alternative approaches to using EHR data from multiple healthcare systems in predicting the outcome of pharmacotherapy for type 2 diabetes mellitus(T2DM). Two of the three approaches, named Selecting Better (SB) and Weighted Average(WA), allowed the data to remain within institutional boundaries by using pre-built prediction models; the third, named Combining Data (CD), aggregated raw patient data into a single dataset. The prediction performance and prediction coverage of the resulting models were compared to single-institution models to help judge the relative value of adding external data and to determine the best method to generate optimal models for clinical decision support. The results showed that models using WA and CD achieved higher prediction performance than single-institution models for common treatment patterns. CD outperformed the other two approaches in prediction coverage, which we defined as the number of treatment patterns predicted with an Area Under Curve of 0.70 or more. We concluded that 1) WA is an effective option for improving prediction performance for common treatment patterns when data cannot be shared across institutional boundaries and 2) CD is the most effective approach when such sharing is possible, especially for increasing the range of treatment patterns that can be predicted to support clinical decision making.

Investigation of time profile of FEV1 across the onset of potential COPD: a retrospective cohort study using medical checkup data in Japan.

This study compared the time profile of FEV1 after COPD diagnosis among rapid decliners, slow decliners, and sustainers in the year of COPD diagnosis. COPD subjects were identified from the annual medical checkup records of Hitachi, Ltd., employees in Japan (April 1998-March 2019). Subjects were categorized into 3 groups (rapid decliner [decrease of FEV1 ≥ 63 mL/year], slow decliner [< 63 and ≥ 31 mL/year], and sustainer [< 31 mL/year]) for 5 years. The time profile of FEV1 was compared using mixed-effects model for 5 years after diagnosis; risk factors of rapid decliner were detected using logistic model/gradient boosting decision tree. Of 1294 eligible subjects, 18.6%, 25.7%, and 55.7% were classified as rapid decliners, slow decliners, and sustainers, respectively. The annual rates of FEV1 decline were similar 3 years before and until COPD diagnosis. The mean FEV1 in rapid decliners was 2.82 ± 0.04 L in year 0 and 2.41 ± 0.05 L in year 5, and in sustainers, it was 2.67 ± 0.02 L and 2.72 ± 0.02 L (year 0, p = 0.0004). In conclusion, FEV1 declined yearly before diagnosis and the time profiles of FEV1 were different in the 3 groups after COPD diagnosis. Therefore, appropriate treatment of the 3 groups with regular lung function tests is necessary to follow FEV1 decline after COPD onset.

Machine Learning Methods for the Diagnosis of Chronic Obstructive Pulmonary Disease in Healthy Subjects: Retrospective Observational Cohort Study.

BACKGROUND: Airflow limitation is a critical physiological feature in chronic obstructive pulmonary disease (COPD), for which long-term exposure to noxious substances, including tobacco smoke, is an established risk. However, not all long-term smokers develop COPD, meaning that other risk factors exist. OBJECTIVE: This study aimed to predict the risk factors for COPD diagnosis using machine learning in an annual medical check-up database. METHODS: In this retrospective observational cohort study (ARTDECO [Analysis of Risk Factors to Detect COPD]), annual medical check-up records for all Hitachi Ltd employees in Japan collected from April 1998 to March 2019 were analyzed. Employees who provided informed consent via an opt-out model were screened and those aged 30 to 75 years without a prior diagnosis of COPD/asthma or a history of cancer were included. The database included clinical measurements (eg, pulmonary function tests) and questionnaire responses. To predict the risk factors for COPD diagnosis within a 3-year period, the Gradient Boosting Decision Tree machine learning (XGBoost) method was applied as a primary approach, with logistic regression as a secondary method. A diagnosis of COPD was made when the ratio of the prebronchodilator forced expiratory volume in 1 second (FEV1) to prebronchodilator forced vital capacity (FVC) was <0.7 during two consecutive examinations. RESULTS: Of the 26,101 individuals screened, 1213 met the exclusion criteria, and thus, 24,815 individuals were included in the analysis. The top 10 predictors for COPD diagnosis were FEV1/FVC, smoking status, allergic symptoms, cough, pack years, hemoglobin A1c, serum albumin, mean corpuscular volume, percent predicted vital capacity, and percent predicted value of FEV1. The areas under the receiver operating characteristic curves of the XGBoost model and the logistic regression model were 0.956 and 0.943, respectively. CONCLUSIONS: Using a machine learning model in this longitudinal database, we identified a number of parameters as risk factors other than smoking exposure or lung function to support general practitioners and occupational health physicians to predict the development of COPD. Further research to confirm our results is warranted, as our analysis involved a database used only in Japan.

Leveraging Artificial Intelligence to Improve Chronic Disease Care: Methods and Application to Pharmacotherapy Decision Support for Type-2 Diabetes Mellitus.

OBJECTIVES: Artificial intelligence (AI), including predictive analytics, has great potential to improve the care of common chronic conditions with high morbidity and mortality. However, there are still many challenges to achieving this vision. The goal of this project was to develop and apply methods for enhancing chronic disease care using AI. METHODS: Using a dataset of 27,904 patients with diabetes, an analytical method was developed and validated for generating a treatment pathway graph which consists of models that predict the likelihood of alternate treatment strategies achieving care goals. An AI-driven clinical decision support system (CDSS) integrated with the electronic health record (EHR) was developed by encapsulating the prediction models in an OpenCDS Web service module and delivering the model outputs through a SMART on FHIR (Substitutable Medical Applications and Reusable Technologies on Fast Healthcare Interoperability Resources) web-based dashboard. This CDSS enables clinicians and patients to review relevant patient parameters, select treatment goals, and review alternate treatment strategies based on prediction results. RESULTS: The proposed analytical method outperformed previous machine-learning algorithms on prediction accuracy. The CDSS was successfully integrated with the Epic EHR at the University of Utah. CONCLUSION: A predictive analytics-based CDSS was developed and successfully integrated with the EHR through standards-based interoperability frameworks. The approach used could potentially be applied to many other chronic conditions to bring AI-driven CDSS to the point of care.

Establishing a multidisciplinary initiative for interoperable electronic health record innovations at an academic medical center.

OBJECTIVE: To establish an enterprise initiative for improving health and health care through interoperable electronic health record (EHR) innovations. MATERIALS AND METHODS: We developed a unifying mission and vision, established multidisciplinary governance, and formulated a strategic plan. Key elements of our strategy include establishing a world-class team; creating shared infrastructure to support individual innovations; developing and implementing innovations with high anticipated impact and a clear path to adoption; incorporating best practices such as the use of Fast Healthcare Interoperability Resources (FHIR) and related interoperability standards; and maximizing synergies across research and operations and with partner organizations. RESULTS: University of Utah Health launched the ReImagine EHR initiative in 2016. Supportive infrastructure developed by the initiative include various FHIR-related tooling and a systematic evaluation framework. More than 10 EHR-integrated digital innovations have been implemented to support preventive care, shared decision-making, chronic disease management, and acute clinical care. Initial evaluations of these innovations have demonstrated positive impact on user satisfaction, provider efficiency, and compliance with evidence-based guidelines. Return on investment has included improvements in care; over $35 million in external grant funding; commercial opportunities; and increased ability to adapt to a changing healthcare landscape. DISCUSSION: Key lessons learned include the value of investing in digital innovation initiatives leveraging FHIR; the importance of supportive infrastructure for accelerating innovation; and the critical role of user-centered design, implementation science, and evaluation. CONCLUSION: EHR-integrated digital innovation initiatives can be key assets for enhancing the EHR user experience, improving patient care, and reducing provider burnout.

A Customized Online Nutrition Guidance System Is Effective for Treating Patients with Nonalcoholic Fatty Liver Disease by Supporting Continuity of Diet Therapy at Home: A Pilot Study.

Continuity is required for diet therapy, but it depends on patients. We examined the utility of a new tool, the customized online nutrition guidance system, in patients with nonalcoholic fatty liver disease (NAFLD). Seven patients plotted their body weight (BW) and marked a customized task card on completion for 90 days on a website. The instructors encouraged them by e-mail. BW, serum transaminase levels, and system usage were evaluated. The results showed that BW and serum alanine aminotransferase levels were significantly lower than at baseline. BW and task visualization as well as encouragement by e-mails were effective in ensuring continuity. Thus, this system is effective in keeping NAFLD patients motivated to continue their diet therapy.

Of the three classifications of healthy lifestyle habits, which one is the most closely associated with the prevention of metabolic syndrome in Japanese?

OBJECTIVE: Metabolic syndrome is associated with a high risk of cardiovascular morbidity and mortality. The predominant cause of metabolic syndrome is an unhealthy lifestyle. Healthy habits are represented by Breslow's 7 healthy practices, Morimoto's 8 items and Ikeda's 6 healthy habits. This study was done to determine which set of healthy habits was most likely to result in a reduced risk of developing the metabolic syndrome. METHODS: From April 1, 2000 through March 31, 2007, 6,765 males and 2,789 females underwent a medical check-up at Jikei University Hospital in Japan. They completed a simple, self-administered lifestyle questionnaire based on the 3 classifications of healthy habits. The responses were divided into 3 groups (poor, moderate and favorable) according to each of the healthy habit criteria. The incidence of metabolic syndrome was defined in participants who were newly diagnosed during the follow-up using Japanese-specific diagnostic criteria. The Kaplan-Meier cumulative 7-year incidence was calculated. Kaplan-Meier curves were compared using the long-rank test adjusted for age. RESULTS: In females, Breslow's, Morimoto's and Ikeda's healthy habits showed significant differences in the incidence between poor and moderate groups, and between poor and favorable groups. In males, a significant difference was observed among the poor, moderate and favorable groups for Ikeda's healthy habits. However, no significant difference was observed for Breslow's healthy practices. Morimoto's items only showed a significant difference between the poor and moderate groups. CONCLUSION: Among the 3 models tested, Ikeda's healthy habits were the most useful for decreasing the risk of metabolic syndrome in Japanese.

Resource management simulator for radiology departments.

We propose a new system to manage modalities in radiology departments. Our system introduces a novel method to accurately estimate some indexes in order to provide an analysis of resource plans within three broad categories: financial affairs, patient satisfaction, and productivity. The main idea is to simulate the operational status of modalities. Results from a preliminary hospital evaluation show that use of the system results in more efficient resource management.

Visualization method for finding critical care factors in variance analysis.

We present a novel visualization method for finding care factors in variance analysis. The analysis has two stages: first stage enables users to extract a significant variance, and second stage enables users to find out a critical care factors of the variance. The analysis has been validated by using synthetically created inpatient care processes. It was found that the method is efficient in improving clinical pathways.

Drug-recommendation system for patients with infectious diseases.

We developed a decision support system that helps doctors select appropriate first-line drugs. The system classifies patients' abilities to protect themselves from infectious diseases as a risk level for infection. In an evaluation of the prototype system, the risk level it determined correlated with the decisions of specialists. The system is very effective and convenient for doctors to use.