Division of Analytics and Informatics in Emergency Medicine (DAIEM)

The Department of Emergency Medicine (DEM) at Yale School of Medicine includes 41 faculty members with 36 clinicians who have responsibility for 82,000 adult patient visits a year at Yale New Haven Hospital and 27,000 visits annually at the satellite ED, the Yale-New Haven Shoreline Medical Center in Guilford, CT. DEM has numerous research studies underway at any given time. EM research is an interdisciplinary field that addresses a plethora of subjects and medical specialties ranging from acute critical care to improving public health. Our faculty are nationally recognized experts in their fields of study, giving tremendous research and training opportunities to advance potential academic careers and research interests. Our work is supported by NIH, AHRQ, PCORI, SAMHSA, CDC, and RWJF, among others.

Division of Analytics and Informatics:

One major aspect of EM research involves a large amount and a wide range of data including clinical data (e.g., patient medical records, drugs, lab tests, etc). With the advent in translational research, there is a growing importance of combining clinical data with molecular data (e.g., genomics and proteomics data) to integrate bench side research into bedside patient care. These data are available through different sources in structured and unstructured formats. With the continued growth of data variety, complexity and quantity, EM research driven by big data calls for advanced informatics and analytics approaches.

The Division of Analytics and Informatics in Emergency Medicine (DAIEM) includes faculty members who have innovative expertise in biomedical informatics, bioinformatics and biostatistics. They have extensive collaborations with researchers and clinicians at Yale, VA, and other institutions. In addition, DAIEM has established a synergistic relationship with Yale Center for Medical Informatics, Yale Center for Clinical Investigation, Yale Center for Analytical Sciences, and Yale Data Coordination Center.

The following are the core DAIEM faculty members:

  • Fuad Abujarad, PhD,MS
  • Brian Biroscak, PhD, MS, MA
  • Cynthia Brandt, MD, MPH
  • Kei-Hoi Cheung, PhD
  • Rachel Dreyer, PhD
  • James Dziura, PhD, MPH
  • Samah Fodeh, PhD
  • Ted Melnick, MD, MHS
  • Michael Pantalon, PhD
  • Andrew Taylor, MD, MHS

ACGME Clinical Informatics Fellowship:

The Clinical Informatics fellowship is a 2-year program that provides ACGME-approved training in all aspects of clinical informatics. The program is administered through the Yale Department of Emergency Medicine. In the first year, the fellow will rotate between the Yale-New Haven Health and Veterans Affairs. Major blocks will be devoted to electronic health records, clinical decision support, databases and data analysis, and quality and safety. Experiential learning will be combined with didactic classes and conferences. The second year is dedicated to advanced learning and project leadership. The fellow will attend the American Medical Informatics Association annual meeting. The program prepares fellows for Clinical Informatics Board examination.

For further information, contact Ted Melnick, MD, MHS, edward.melnick@yale.edu (program director) or Dena Blancato, denamarie.blancato@yale.edu (program coordinator).

VIC : mHealth Tool for the Informed Consent Process

Informed Consent (IC) is a process in which patients are educated about the benefits/costs of medical procedures and clinical research trials that they choose to participate in. It allows for a more ethical conduct of medical treatment and research. The Joint Commission (2007) reported that, “among patients who sign an IC form, 44% do not understand the nature of the procedure to be performed, and 60-70% did not read or understand the information contained in the form.” Past research has found that paper and electronic IC forms do not guarantee patient comprehension and satisfaction. To minimize future costs/risks for the hospital and patient, patients must be able to understand the IC form contents so they can be informed about the purpose and functionality of medical procedures.

VIC.png

The purpose of this study was to develop and evaluate the usability of the Patient Centered Virtual Multimedia Interactive Informed Consent tool (abbreviated VIC). The overall study was a mixed methods observational design of qualitative and quantitative methods with two specific aims.

Aim 1: To develop, test and refine the VIC mHealth tool via User-Centered Design (UCD) methodology and User Experience (UX) evaluations. The VIC mHealth tool has a reusable infrastructure in integrating patient-centered IC processes to improve patient care and satisfaction. Front-end short interviews and a focus group with patients, researchers and IRB members were used to help make a preliminary prototype of VIC. In before-launch evaluation, there were 10 representative asthma patients participating in one-on-one UX sessions to compile patient performance and satisfaction data. In after-launch evaluation, there were 6 representative asthma patients providing their opinions and attitudes toward VIC in an effort to understand VIC’s generalizability.

Aim 2: To evaluate VIC usability in comparison to a coordinator-delivered consent form in an ongoing clinical research study, Yale Center for Asthma and Airway Disease Mechanisms and Mediators of Chronic Lung Disease Study (YCAAD GenEx). 100 eligible asthma patients were randomized to receive either standard paper consent (SIC) or consent via VIC. The VIC process was evaluated in perceived understanding, satisfaction and voluntariness of consent. VIC provided a ‘teach-back’ process, automated readability evaluation, Internet access to the consent, retrievable electronic record of IC, electronic signature, and integration with electronic health record.

In regards to progress, the average SUS (system usability scale) normalized score for all participants was 90th percentile, which is above the industry benchmark of 68th percentile.  The participants unanimously recommended its use in future clinical practice and treatments. Possible future directions include assessing the VIC tool capabilities with different chronic diseases and seeing if patients continue to report statistically significant improvement in perceived understanding and satisfaction. Potential modifications include improving the patient-centered infrastructure of the mHealth tool by using a larger and more diverse sample size. The VIC mhealth tool is a huge first step in developing a functional, highly satisfactory IC tool that can become more of a staple in the healthcare/research space and help patients better understand their own healthcare choices and decisions.

OverdosePreventionEfforts.png

Strengthening Prescription Drug Overdose Prevention Efforts is a CDC program that helps states combat the ongoing prescription drug overdose epidemic. The purpose of Prevention for States is to provide state health departments with resources and support needed to advance interventions for preventing prescription drug overdoses. In March 2016, Connecticut joined 13 other states to receive this funding making it a total of 29 states throughout the United States. The Connecticut Department of Public Health, in collaboration with the Departments of Mental Health Addiction Services (DMHAS) and Children and Families (DCF) and other stakeholders in the state, are currently working with Yale’s DAIEM to evaluate the Connecticut Strategic Plan on Addiction and Overdose. DAIEM faculty are using system dynamics modeling—i.e., the use of causal diagrams and computer simulation models to hypothesize, test, and refine explanations of systems change—to evaluate the impact of Connecticut’s prescription drug overdose prevention efforts.



ImpactsandImpactPathwaysofFood.png

A Research Strategy to Understand the Impacts and Impact Pathways of Food, Agriculture and Nutrition Policies in Feed the Future (FtF) Countries is an effort of a multi-institution consortium to help advance sustainable development goals. The Department of Emergency Medicine at Yale School of Medicine is a member of this USAID-funded consortium. DAIEM faculty are exploring the interrelationship between climate change, food security, and gender-based violence. In addition to disparities along the gender dimension, this project is focused on addressing poverty as a root cause of food insecurity and violence—both exacerbated by climate change. DAIEM faculty are using Leximancer computer software, which allows its users to conduct quantitative analysis using a machine learning technique, to perform a scoping review of published evidence regarding the interrelationship between climate change, food security, and gender-based violence.


Dr. Brandt’s research focuses on informatics in clinical research and health services research. She currently directs several health services research projects that focus on the use of data from electronic health records.

Dr. Brandt is Co-Director of the VA Postdoctoral fellowship in informatics and the Advanced Fellowship in Women’s Health. She is also Co-PI on a HRS&D CREATE musculoskeletal pain cohort project that this research grant is leveraging for the EHR data, and another EHR-based grant studying gender differences in veterans returning from Iraq and Afghanistan. In addition I oversee research data management at VACHS and Yale, where we support REDCap, Oncore and other systems.

Dr. Brandt’s other current research is interdisciplinary and focuses on issues related to the design, development and use of informatics tools to extract and use clinical data from electronic health records for health services research and quality improvement. Dr. Brandt is the PI and Co-Investigator on several health services research informatics projects working on the development and the use of open-source natural language processing (NLP), informatics tools for information retrieval and information extraction for electronic medical record free-text data.

In addition, Dr. Brandt’s research involves developing multi-view learning methods to enhance mining several sources of data. Applying such methods can help classifying and clustering patients with several health care conditions, especially when utilizing different aspects about these patients such as their diagnostic imaging data, genomic data or clinical text data.

Database, Ontology, Natural Language Processing

Advanced informatics infrastructure and computational approaches are critical to today’s big-data-driven biomedical research ranging from the high throughput omics research to patients’ electronic health records. These informatics/computational components include databases/data warehouses, ontologies, and natural language processing (NLP). Below are some of the studies conducted at Yale and VA which involve these components.

core DAIEM faculty image 1.png

Neuroproteomics database (YPED): The Yale/NIDA Neuroproteomics Center brings a number of Yale programs in proteomics and signal transduction in the brain together with neuroscientists from nine other institutions across the U.S. to identify adaptive changes in protein signaling that occur in response to substances of abuse. The Yale Protein Expression Database (YPED) is an open-source proteomics suite and database designed to address the storage, retrieval, and integrated analysis of high throughput proteomic and small molecule analyses. For a full list of publications generated by the Center, please see the following: http://medicine.yale.edu/keck/nida/publications/publications.aspx


Data/metadata standards: As part of the data/metadata standards effort of the Human Immunology Project Consortium (HIPC) (http://www.immuneprofiling.org/hipc/page/show?pg=publications) and Extracellular RNA Communication Consortium (ERCC) (http://exrna.org/publications/), the elements of metadata/data templates are mapped to standard terms defined in ontologies. Such ontology term mappings provide common concept definitions and relationships for enabling semantic data integration and queries across different laboratories. In addition to HIPC and ERCC, the ontology term mapping approach has been extended to support the metadata activities carried out by the Center for Expanded Data Annotation and Retrieval (CEDAR) (http://metadatacenter.org/research/publications) as part of the Big Data to Knowledge (BD2K) initiative.

Data warehouse and NLP projects: The Department of Veterans Affairs (VA) and the Yale School of Medicine have the mission to provide a high-performance business intelligence infrastructure through standardization, consolidation and streamlining of clinical data systems. Toward this end, the VA has built the Corporate Data Warehouse (CDW) and EPIC is currently used by the Yale-New Haven Hospital to store and access electronic medical records. The data warehouse systems contain a large amount of structured data (e.g., diagnoses, medications and lab tests) and unstructured data (e.g., various types of clinical notes). A wide variety of VA and Yale projects require retrieval and extraction of these data using ontologies and NLP methods to support different kinds of clinical research projects. Among these projects are: the million veteran project (MVP) including the genetics of functional disability in schizophrenia and bipolar illness (https://clinicaltrials.gov/ct2/show/NCT01149551), the cancer tracking system (CCTS), psychogenic nonepileptic seizure research, and veteran aging cohort study (VACS) (http://medicine.yale.edu/intmed/vacs/resources/).

Coronary heart disease (CHD) is the leading cause of death in women, yet nonetheless has been defined primarily a ‘man’s disorder’, and therefore not an important health concern for women. This lack of education has resulted in many women not being appropriately informed of their cardiovascular risks and consequently in health care providers underestimating this threat as compared with men. Consequently, evidence-based clinical standards and guidelines have been created based on male pathophysiology and outcomes. Clear disparities exist in regards to the presentation, diagnosis, and management of women with CHD, leaving many questions unanswered.

Dr Dreyer’s principal research interests focus on understanding sex and gender- specific issues in CHD and the cause of the poorer outcomes in women. To do so, she applies her unique training in mixed methods and health services/outcomes research to the analysis of observational, medical claims, and administrative data. Dr Dreyer’s research to date has demonstrated significant sex and gender based disparities in the diagnosis and management of acute myocardial infarction (AMI) in women across the continuum of care, including pre-AMI care, in-hospital treatment and follow-up care. For example, women report significantly poorer physical/mental functioning, more angina and poorer quality of life up to a year after their AMI; as well as higher rates of depression and stress, compared with men. In addition, women have higher rates of hospital readmission, and are less likely to return to work.

Dr Dreyer’s current work focuses on bridging outcomes research to implementation science, with the goal of developing mobile health (mHealth) applications to assist patients, particularly women, in recovery after AMI, for example by creating personalized secondary prevention. More broadly, she is interested in the application of big data analytics, and the potential for gendered innovations in design.

Data Coordination for Multicenter Studies

The Yale Data Coordinating Center (YDCC) is a partnership between the YCAS, Emergency Medicine, the Yale Program on Aging (POA) and the Yale Center for Medical Informatics (YCMI). It is composed of faculty from the School of Medicine and School of Public Health with expertise in biostatistics, epidemiology, clinical trials and informatics along with a highly trained technical staff skilled in systems programming, data management, data analysis and statistical programming. Our group has experience designing and conducting single and multicenter clinical trials, longitudinal cohort studies, case control studies and other observational studies.

Clinical Decision Support

DecisionMaking.png

Overuse of advanced diagnostic imaging in injured patients in the emergency department has increased dramatically correlating to increased health care costs, exposure to ionizing radiation, and length of stay without a measurable improvement in patient outcomes. Validated, evidence-based guidelines and computerized clinical decision support are promising; however, there are many challenges to implementation including awareness and adherence to guidelines and usefulness, usability, and integration of decision support into clinical workflow.  


Dr. Melnick is now three years into a K08 award with the Agency for Healthcare Research and Quality. This study aims to pilot an innovative clinical decision support design process that produces decision support for patients and their providers that is patient-centered, useful, usable and promotes shared decision-making for the management of minor head injury in the emergency department.

  1. Melnick ER, Szlezak CM, Bentley SK, Dziura JD, Kotlyar S, Post LA. Overuse of CT for mild traumatic brain injury. The Joint Commission Journal on Quality and Patient Safety. 2012;38:483-489.
  2. Melnick ER, Lopez K, Hess EP, Abujarad F, Brandt CA, Shiffman RN, Post LA. Back to the bedside: Developing a bedside aid for concussion & brain injury decisions in the emergency department. eGEMs (Generating Evidence & Methods to improve patient outcomes).2015;3(2):1136.
  3. Melnick ER. How to make less more: Empathy can fill the gap left from reducing unnecessary care. BMJ. 2015;351:h5831.
  4. Melnick ER, Shafer K, Rodulfo N, Shi J, Hess EP, Wears RL, Qureshi RA, Post LA. Understanding Overuse Of CT For Minor Head Injury In The ED: A Triangulated Qualitative Study. Academic Emergency Medicine. 2015;22:1474-1483.
  5. Melnick ER, O’Brien EGJ, Kovalerchik O, Fleischman W, Venkatesh AK, Taylor RA. The association between physician characteristics and variation in imaging use. Academic Emergency Medicine. Jun 25. doi: 10.1111/acem.13017. [Epub ahead of print]
  6. Melnick ER, Probst MA, Schoenfeld E, Collins SP, Breslin M, Walsh C, Kuppermann N, Dunn P, Abella BS, Boatright D, Hess EP. Development and testing of shared decision-making interventions for use in emergency care: A research agenda. Academic Emergency Medicine. [in press: AEMJ-16-389.R1]

Machine Learning

MachineLearning.png

Traumatic brain injury (TBI) is a critical public health and socio-economic problem and a major cause of death and lifelong disability. There are 1.7 million TBIs and 52,000 deaths from TBI annually in the United States--600,000 of those injured are children resulting in more than 6000 annual pediatric deaths. In addition, there are 5.3 million Americans who are living with TBI-related disability. This heterogeneous disease remains poorly understood with very limited treatment options available. TBI in children is particularly devastating given the possibility of lifelong impairment in previously healthy individuals.


Aim 1. To identify proteins in the blood following concussion in adolescent athletes that are significantly and consistently altered compared with preseason samples from the same individuals.
Aim 2. To identify proteins in the blood in the acute phase after concussion in athletes that develop post-concussive syndrome that are significantly and consistently altered compared with athletes with concussion that do not develop post-concussive syndrome.
Aim 3. To develop a supervised machine learning algorithm based on proteomic data and varied downstream disease states following concussion that is predictive of development of post-concussive syndrome.

 Two recent machine learning papers using random forest models are:

  1. Melnick ER, O’Brien EGJ, Kovalerchik O, Fleischman W, Venkatesh AK, Taylor RA. The association between physician characteristics and variation in imaging use. Academic Emergency Medicine. Jun 25. doi: 10.1111/acem.13017. [Epub ahead of print]
  2. Taylor RA, Pare J, Venkatesh AK, Mowafi H, Melnick ER, Fleischman W, Hall MK. Prediction of In-hospital Mortality in Emergency Department Patients with Sepsis: A Local Big Data Driven, Machine Learning Approach. Academic Emergency Medicine. 2015 Dec 17. doi: 10.1111/acem.12876. [Epub ahead of print]

Concussion Biomarker Discovery for Prognosis in Adolescent Athletes (COMPETE)
This biomarker discovery project is in a pilot phase now. The objective of this study is to identify proteins in the blood following concussion that are altered. We hypothesize that proteins not normally present in significant quantities in the blood will appear and/or become elevated following concussion. We will use a random forest supervised machine learning model using feature selection to determine which proteins have the strongest correlation with concussion severity and time to resolution of concussive symptoms.

Pantalon.png

Dr. Pantalon conducts research on screening and brief motivational interventions (e.g., the Brief Negotiation Interview or BNI) aimed at reducing health-risk behaviors, including addiction treatment non-engagement, opioid and alcohol use, smoking, HIV risk behavior, poor management of asthma, and failure to report elder abuse. His work also focuses of the development and testing of web-based methods for both delivering and training healthcare professionals in these approaches, with an emphasis on intervention fidelity using the Brief Negotiation Adherence Scale (BAS).


Research Staff

Fuad Abujarad, PhD, MSc

Assistant Professor of Emergency Medicine

Brian Joseph Biroscak, PhD, MS, MA

Assistant Professor of Emergency Medicine

Cynthia A Brandt, MD, MPH

Professor of Emergency Medicine and of Anesthesiology

Kei-Hoi Cheung, PhD

Professor of Emergency Medicine

Rachel Dreyer, PhD

Associate Research Scientist

James David Dziura, MPH, PhD

Professor of Emergency Medicine

Ted Melnick, MD, MHS

Assistant Professor of Emergency Medicine

Michael V Pantalon, PhD

Senior Research Scientist in Emergency Medicine

Richard Andrew Taylor, MD, MHS

Assistant Professor of Emergency Medicine