OHDSI: Enterprise Research Data System at Montefiore-Einstein
As of July 1 2020, all our investigators have access to a modern data platform to use clinical and health data from Montefiore patients. The system is integrated with the Epic EMR system, Velos (our CTMS), and the Montefiore Enterprise Data Warehouse (EDW). Clinical data from all Montefiore sites which use Epic will become available for research, advanced analytics, informatics, and data science by Montefiore- Einstein users. Currently these include MMC inpatient and outpatient sites, White Plains Hospital and ambulatory clinics, but will soon expand to all MHS sites as the Epic is rolled-out across MHS locations.
The Center for Health Data Innovations (CHDI) has implemented an open source technology developed by an international coalition of health data informaticians, data scientists, and clinical researchers called Observational Health Data Sciences and Informatics (OHDSI, pronounced “odyssey”). The OHDSI community shares rigorous research methodologies, advanced analytic methods, generalizable models, and software tools that will be available for open academic use at Montefiore and Einstein. Over the past several years using OHDSI for clinical data has become an agreed-upon best practice, and a de-facto standard for real-world evidence generation across universities, pharma, and healthcare systems. The NIH and other research funding agencies recognize OHDSI, and preferentially support co llaborations, research projects, and initiatives that are based on OHDSI standards. To learn more about OHDSI you can visit http://OHDSI.org. The Einstein-Montefiore CHDI has been using OHDSI tools in support of AHRQ, NIH, and PCORI grants since 2015 and is now expanding its implementation MHS-wide to support the broader investigative, data science, and informatics communities.
This first phase of OHDSI services became available July 1, 2020, and will incrementally expand throughout the summer and fall 2020 to support a wide spectrum of clinical research, from retrospective and prospective observational studies, to comparative outcomes research, quality of care, patient level predictive modeling, pragmatic clinical trials, operational and clinical real-time monitoring, just-in-time signal detection, and situational awareness systems for decision support.
To obtain access to data for your research, you must submit a protocol to the Einstein IRB (please see first FAQ below for details). If you have an existing approved project, we encourage you to update your IRB protocols to ensure uninterrupted access to the systems and its services.
What types of Montefiore data will be available through CHDI?
ATLAS is a user-friendly, web-based application that interactively supports a wide variety of analyses which requires patient level clinical data (self-service analytics). For example, ATLAS enables users to define patient cohorts (for example, Covid-19 patients with history of ACE inhibitor use, with cardiovascular comorbidities), compare baseline characteristics and differences between cohorts and/or subgroups within cohorts, measure incidence rates, define effect sizes and estimates, study clinical pathways within cohorts, or develop patient level predictions without ever having to write any R, SPSS, or SAS codes.
The CHDI instance of ATLAS enables feasibility or prep-for-research studies based on Montefiore clinical data and can be used to generate dynamically updating registries of Montefiore patients for longitudinal studies, pathway analysis, and outcomes research, through an easy to use interactive GUI. Learn more about ATLAS with video tutorials. You can review “The Book of OHDSI ” for a more complete description of how ATLAS supports cohort characterization, Clinical pathway analysis, incidence rate analysis, effect size estimates, and patient level predictions.
The Observational Medical Outcomes Partnership (OMOP)
Common Data Model
The OMOP Common Data Model reconciles the differences between disparate databases (e.g. Epic Clarity, EDW, Administrative Claims) by adopting a common representation of various terminologies, vocabularies, and coding schemes practiced across multiple systems and communities (ICD 10, ICD 9, CPT, LOIC, SNOMEDCT, RXNorm, NDC, and many others). For example, the Montefiore instance of OMOP-CDM uniformly represents EHR data from Epic Clarity, Montefiore EDW, and Epic Chronicle, allowing users to effectively perform systematic analyses on data by using a large library of statistical tests that are validated on OMOP-CDM, to easily generate evidence or validate hypothesis based on multi-source clinical data from Montefiore. ATLAS uses the OMOP-CDM as its data source. The standard nature of OMOP CDM enables research groups, and institutions to easily collaborate by sharing their data, algorithms, and methods since they all adopt the same data language for collaboration. In effect, this has made the OMOP-CDM a standard for research networking and a critical infrastructure for grant applications for large scale data driven research. As of July 1, 2020, Einstein and Montefiore has participated in 3 NIH funded national, regional, and local research consortiums and 1 PCORI funded research network all based on OMOP-CDM. Read more about the Common Data Model (CDM). You can also find more information about other OHDSI tools such as HADES, ACHILLES, ETNA and other components that are available through the CHDI instance at the OHDSI.org website.
Phase 1 Data Availability
Phase 1 data has become available from July 1, 2020 and will continue to expand until end of September 2020. The phase 1 data will be updated daily. CHDI will continue adding new data types from varieties of clinical systems and datasets throughout phase 1. Below is a summary of what is available as of July 1, and what will be incrementally added to the OMOP-CDM and ATLAS later on.
Currently Available Data
- Clinical data from Epic: Patient level demographics, visits (inpatient, outpatient, telehealth, etc.), care-sites, providers, conditions, diagnoses, clinical observations, medications, and laboratory tests.
- Data from select Epic flow-sheets: This will be incrementally expanded based on priorities and projects. Please communicate with the CHDI if you are interested in extracting Flowsheet data for research and analytics use.
- Clinical Notes: All clinical notes from admission to discharge, including radiology and pathology reports, consultations, and chief complaints and reasons for admission.
Available By the End of Phase 1
- Raw data from Devices: Starting with ventilators and respiratory devices, and gradually expanding to other monitoring devices, and waveform data (ECGs).
- Data from Montefiore PACS systems: Imaging studies (X-ray, MRI, CT, etc.) will be made available for cohort definition, chart review, image processing, and data science pipelines through the phase 1. We are developing a generic method for de-identification of image files that will accompany image extraction features for large scale download and use of images for research.
Phase 2 will start in October and will continue throughout fall 2020, and will include:
- Cost of care data (encounter and service level)
- Payer Plan Period data
- Administrative claims
- NLP output for all clinical notes.
- Fields from disease registries (eg, tumor staging data from the Cancer Registry)
Phase 3 will go live in December 2020 in order to implement real-time updating of data. This will enable the Montefiore-Einstein community to engage in real-time monitoring, situational awareness, just-in-time signal detection, and real-time decision support projects, as well as pragmatic clinical trials integrated into the delivery of care process MHS-wide, using timely and high quality analytics.
How to Access
Prior to accessing ATLAS, please review the FAQs below. To request access to ATLAS, use our portal to submit your request. Your application will be reviewed and proper access will be granted. For questions and inquiries please contact CHDI or email email@example.com.
Frequently Asked Questions
Please submit your request to access OHDSI platform through our portal. You will need a working Einstein-Montefiore username and password (same username and password you use to login to your Montefiore or Einstein email), and have submitted an IRB protocol. At this point you will be able to login for cohort definition, characterization or feasibility studies while your IRB protocol is still under review, but would only be able to download the list of patients after the IRB approves access to identified data. In different stages of your IRB protocol review process, you may have different levels of access to the system functionality. In order to export individual level patient data from ATLAS or OMOP-CDM you will need appropriate IRB approvals. You will receive a link to access ATLAS once you are registered as an OHDSI user. We recommend that you bookmark the link in your favorite browser for easy access and future use.
To view the basic Atlas Platform you can log in using your Einstein-Montefiore username and password (same username and password you use to login to your Montefiore or Einstein email). Montefiore’s Atlas instance: https://chdi.montefiore.org/atlas/. You will not be able to see any data but can see the basic structure.
View Deidentified Data on ATLAS
Please submit your request to access OHDSI platform through our portal.
The Atlas Administrator will activate your account and reply to your request with detailed instructions
View Identified Data on ATLAS
You need an approved IRB Protocol to view Identified data
Please submit your request to access OHDSI platform through our portal.
The Atlas Administrator will activate your account and reply to your request with detailed instructions for uploading your protocol and signing your user agreement.
Since most CHDI resources are currently dedicated to developing and deploying the OHDSI data and technology infrastructure, one-on-one training may not be immediately available for all users and projects. However, the ATLAS web-application is intuitive and easy to use. There is a wealth of information, tutorials, and “the Book of OHDSI” to learn more, as below:
You must use a device on the Montefiore-Einstein network — you must either be on site at Montefiore or Einstein or remote into your onsite computer. We are working with Montefiore IT to make the platform accessible from outside the network.
You must use Google Chrome Browser for full functionality.
There is no charge or fee to use Atlas.
Yes. ATLAS supports patient level data export, under an approved IRB protocol. ATLAS will provide different levels of identification or de-identification (fully de-identified, limited, or identified) for data export, as long as you have an approved IRB protocol associated with the export. If your IRB approval permits, ATLAS will be able to re-identify (via an automated honest-broker process) subsets of patients in a de-identified cohort as well (Phase 2 functionality). If your IRB protocol is not active, or you are not listed as key personnel in an approved protocol, you won’t be able to access or export patient level clinical data, but you may be able to continue working on your cohort definitions, prep for research, study design, and analytics that only involve aggregate level information based on de-identified data. The data export functionality of ATLAS is under active development and more refined functionality and advanced features will be incrementally provided throughout the phase 1 implementation period (July – September 2020).
Yes. This is a phase 1 functionality and will be developed and refined throughout the summer. Under an approved IRB protocol, you will be able to link your own cohorts and patient registries to ATLAS for analytics or for data extraction. We recommend that you maintain your patient registries in the Einstein-Montefiore Research Electronic Data Capture ( REDCap ) system to ensure security and HIPAA compliance. REDCap will be integrated to ATLAS to automatically support clinical data extraction and analytics on REDCap based registries, or cohorts extracted from other systems such as EDW or CLG. You will be able to schedule and automate population of your patient registries through upcoming REDCap integration. If you are recruiting patients into Velos Clinical Trial Management System (Velos), you will be able to link your trials to ATLAS for access to clinical data (lab results, diagnoses, etc.) and prospective analysis automatically (in Phase 2). Uploads from Excel or flat files (.csv) are not desirable but will be supported. Please consider use of REDCap for security and HIPAA compliance.
Yes. But this is a phase 2 functionality. The CHDI has developed an automated patient communication, consenting and recruitment process that is currently linked to REDCap. During the phase 2, we will integrate the patient communication, consenting and recruitment functionality to ATLAS cohorts. You will be able to use your ATLAS defined cohorts to invoke the automated process to obtain permission to contact and enroll patients in your trials from patient’s PCP, send automated text, voicemail, and email messages to approved candidates, compile a daily list of approved candidates in each outpatient clinic for consenting or intercepting patients face to face, or use an e-Consenting service to enroll patients online. This process is fully automated and will be linked to Velos for automated enrollment if you are recruiting patient for a clinical trial. Please contact CHDI if you have an immediate need to use the automated patient communication, consent, and recruitment system.
ATLAS will support you with interactive cohort definitions, cohort characterizations (comparing baseline features and characteristics of multiple cohorts, and subgroups within cohorts), clinical pathways (course of clinical events within a cohort), incidence rate analysis, effect size estimations, and patient level predictions out of the box and interactively. There are hundreds of automated measures for temporal analysis of cohorts, and a large library of methods available to implement rigorous research studies using interactive tools.
The CHDI implementation of OHDSI provides a library of method (implemented as R packages) that can be used together to develop a complete observational study, starting from data in the OMOP-CDM, and resulting in estimates and supporting statistics, figures, and tables. The library provides advanced standardized methods to implement rigorous population characterization, population-level effect estimation, patient-level prediction s and more. These methods support best practices for use of observational data and observational study design such as transparency, reproducibility, measuring the operating characteristics of methods and subsequent empirical calibration of estimates produced.
You will be able to share and use queries and analytic methods with your collaborators inside and outside of Montefiore-Einstein. The OHDSI community maintains a large library of validated and reusable cohort definitions, and research methods that are available openly for sharing and re-use by OHDSI member organizations. You will be able to link your patient registries and cohorts to ATLAS for data export, analytics and data science. If you are a REDCap user or conduct a clinical trial at Montefiore, you will be able to use Velos and REDCap to access and use patient level clinical data (under an approved IRB protocol). You will be able to participate in multi-institutional, data driven research collaborations nation-wide. Currently there are several regional and national consortiums comprised of academic medical centers and pharma collaborating on large-scale clinical research. Montefiore-Einstein participates in 3 of such consortia nation-wide for Covid-19 research, providing opportunities for research for our researchers. Please contact CHDI, if you are interested in learning more about data availability and research opportunities from these consortiums.
The CHDI has implemented the Data Quality Dashboard (DQD), an OHDSI tool that applies a systematic data quality assessment process to underlying OMOP-CDM data. The DQD evaluates the data table-by-table and field-by-field to quantify the number of records that do not conform to the given data quality specifications. In all, over 1,500 checks are performed. For each check, the result is reported as PASS or FAIL depending on if the percentage of violating rows is above a certain threshold value. Furthermore, the CHDI has implemented its own data quality monitoring system (QualiPy) that uses artificial intelligence and machine learning for aberration and anomaly detection, to sense subtle shifts in quality and completeness of data and alert users and system administrators for change over time in quality of data. Data quality is a moving target and requires diligent and continuous monitoring and corrective measures. Combination of our technology tools and collaborative work with many MIT, Montefiore, and Einstein teams are brought to bear on providing the highest quality data for our users, and a transparent and objective view of the underlying issues.
CHDI is collaborating with the departments to train super-users and support staff in each department that can provide departmental services for use of OHDSI tools. CHDI and the department of Epidemiology and Population Health (through Mindy Ginsberg) are planning to provide training and analytic support services for projects that need further assistance or have time-sensitive milestones. But remember, ATLAS is a do-it-yourself tool. Please review the links below for a more in depth documentation and resources.
There are 2 ways to import your cohorts from CLG to ATLAS, but not automatically.
- Re-define your CLG cohorts in ATLAS: ATLAS provides a wide range of cohort definition functionalities, complete with an array of robust time-based feature analysis and feature engineering tools to define the clinical events and the relationships between them. While the nomenclature and GUIs are different, the cohort definition in both systems is designed to provide similar functions.
- Export your cohort of patients from CLG and import as a user-defined cohort in ATLAS.
The Montefiore Einstein Center for Health Data Innovations (CHDI) is a multidisciplinary unit directed by Parsa Mirhaji MD, PhD. The Center supports Montefiore’s learning health system and builds bridges to research and data technologies that are integrated with clinical practice. CHDI’s education and training programs are based on collaboration, patient engagement, and opportunities for trainees to participate in research. CHDI provides the expertise, methodologies, and infrastructure to leverage multi-source, heterogeneous, and linked data for evidence generation for clinical decision making, predictive analytics, and translational research.
Generate Data. Inform Your Research. Make Better Health Decisions.