Skip to main content



Agenda of the
Electronic Health Record Informatics Workshop
29 th Human-Computer Interaction Lab Symposium
University of Maryland
Wednesday, May 23, 2012

Workshop schedule, participants, talk titles, and abstracts are below.
Click here to return to main Electronic Health Record Informatics Workshop page.

Draft Agenda

8.15am Registration & Breakfast
9am Demos & Posters
10.30am Welcome and Greeting Catherine Plaisant, Associate Director of Research, HCIL, UMD
10.45am Introduction Ben Shneiderman, Professor, Department of Computer Science, UMD
Session 1: Visualization
11am Visualizing Patterns of Asthma Medication Prescribing in the Military Health System using EventFlow Tamra Meyer, Megan Monroe, Jeff Millstein, Suji Xie, Trinka S Coster -- Pharmacovigilance Center, Office of the Surgeon General, Department of the Army; University of Maryland, Computer Science and HCIL; Oracle Health Sciences

Background: FDA and DoD are interested in detecting misprescribing of long-acting beta-agonists (LABAs) in asthmatics within the Military Health System (MHS). Visualizing the patterns of asthma medication use surrounding a LABA prescription is a quick way to detect potential LABA misprescribing for further evaluation. Methods: We selected a random sample of 100 asthma patients under age 65 with a new LABA prescription from January 1, 2006-January 1, 2011 in MHS healthcare claims. Visualization was conducted in EventFlow; software developed by the University of Maryland Institute for Advanced Computer Studies' Human Computer Interaction Laboratory to display time-point and interval data. All asthma medication prescriptions for these individuals were extracted from MHS claims and grouped into useful categories including LABAs, short-acting beta agonists (SABAs), and inhaled corticosteroids (ICSs). The prescription days supply was added to the dispense date -1 to determine the prescription interval. Individuals that shared the same sequence of medications were combined and the average interval times were displayed.

Results: When we aligned asthma medication intervals on each person's first LABA/ICS prescription, we noted patterns that might be sub-optimal. For example, about 40% of the population started a LABA/ICS without a previous asthma medication which is sub-optimal except in a severe asthma exacerbation. Most individuals had several asthma medications after their first LABA/ICS which identifies a group of patients that may benefit from monitoring for step-down therapy intervention. A few individuals had a pattern of SABA use that suggested the need for earlier LABA intervention.

Conclusion: Use of EventFlow was a quick, easy, and powerful tool for visualizing possible patterns of sub-optimal LABA use that can be further evaluated in detail elsewhere. The ability to include larger samples will be necessary to maximize the utility of EventFlow for research and surveillance studies.

11.20am Visual Analytics for Tracking Disease Progression in Electronic Health Records Adam Perer -- IBM

The growing use of Electronic Health Records yields a massive amount of long-term data about patients, diseases, and providers. In order to make sense of this data, we are developing many healthcare analytics and visualization technologies to help clinicians understand and analyze these longitudinal records. In this talk, I will focus on our recent work in designing visual interfaces to help the analysis of disease progression to potentially improve disease diagnoses and treatments.

My talk will use heart failure as a motivating example for our analytics. Congestive heart failure occurs when the heart cannot supply the necessary blood flow to meet the needs of the body. This condition is potentially fatal and affects about 2% of adults in developed countries. Despite its widespread occurrence, the disease is difficult to manage and there are currently no systematic diagnostic criteria. I will highlight how our analytics aid our medical collaborators to better understand the patterns of patient symptoms and how their order of onset correlates with patient outcome.

11.40am Integrating Cognitive Science and Information Visualization in Modality Management Paul Nagy -- Johns Hopkins University

A graphical web tool was developed to help technologists and supervisors visualize the complexities of orchestration around operating an imaging modality in a busy radiology department. The real time tool provided situational awareness and a system flow perspective that added in root cause analysis. The system was used to perform continuous performance improvement interventions to reduce delays and bottlenecks. The system was developed on Ruby on Rails with an image library, Chartdirector, along with javascript to provide interactivity.

12pm Real-time Radiology Workflow Monitoring Tool for Overnight Residents with Flot Michael Cohen -- Johns Hopkins University

During the overnight hours, residents in the Radiology Department must read and review all cases that come into the hospital, no matter the modality. To relieve workload stress and protect residents from potentially harmful diagnostic errors, we created a real-time one-off radiology monitoring tool.

This monitoring tool displays the list of cases in the resident’s worklist in a stacked bar chart by modality and a Gantt chart to visualize the time each case is in the queue before its preliminary signoff by the resident. The Gantt chart is color coded by modality. Such a tool can be utilized to visualize the need for additional readers based on the amount of cases in the queue and the possibility of strain on overnight residents. The tool provides a method for users to blog about the nightly events and associates them with their respective cases. This provides additional understanding when retrospectively reviewing and attempting to determine why caseload may have increased or decreased during specific overnight hours. Totals of cases read by modality and overall, relative value units completed and unread cases (at that time) are calculated.

We chose the open-source Flot Graphical Library to easily graph data with the HTML5 Canvas element. Using this library we customized the graph to show our specific time of interest, 8pm to 8am, as well as color coding the data by modality. The Flot Library has additional plug-ins for panning and zooming interactivity that we used to focus on a specific area of both charts simultaneously for additional analysis. We are able to get access to the real-time data feed of ADT, ORU and ORM HL7 messages through a Software Development Kit (SDK) provided by the vendor framework we have deployed at Johns Hopkins.

Multiple one-on-one design sessions between the web developer and a select group of radiology residents were used to maximize the usability of the tool by visualizing the appropriate data in the most readable form. Once mockups were complete, a prototype was created and enhanced. The final tool is based on the enhanced prototype and further advanced with real-time data.

12.20pm Lunch
Session 2: EHR Design
1.40pm Electronic Health Records and the Clinical Narrative Philip Resnik -- UMD

In order for electronic health records to live up to their full potential, all that information about patients -- their symptoms, diagnoses, allergic reactions, medical backgrounds, family histories, and the rest -- must be standardized, structured, and easy to manipulate. The most obvious way to get clinical information into that form is to structure the way that doctors create the medical record in the first place. As a result, there is considerable momentum in the direction of abandoning traditional clinical dictation and turning the medical documentation process into a highly structured activity driven by pull-down menus, checkboxes, templates, and restricted vocabularies. In this presentation I look at the potential impact of this restructuring on health records and the clinical narrative, arguing that we cannot afford to reduce unrestricted clinician language to textboxed afterthoughts in the design of EHRs. I will illustrate how the necessary movement toward structure be driven from the unstructured side, using natural language processing as a key technology to translate from unstructed to structured representations.

2pm Development of a Model Electronic Health Record Format for Children P. Kenyon Crowley, MBA, MS -- UMD Center of Excellence in HIT Research & CHIDS

Electronic health records (EHRs), when fully implemented well, are comprehensive systems that support the health record-related needs clinicians and all their patients. Unfortunately, for a variety of reasons, EHR systems often do not adequately support the provision of health care to children. The project goal is to develop a Model EHR Format for children, which specifies: A minimum set of data elements; Applicable data standards; Usability; Functionality; and, Interoperability requirements. Then, to demonstrate that it can be readily used, and package it in a way that facilitates broad incorporation into EHR systems. The project output will help facilitate improved care coordination and management throughout the children's health ecosystem. Support provided by AHRQ.

2.20pm A Systematic Yet Flexible Systems Analysis of Decision Entry Eliz Markowitz -- University of Texas

Clinical decision support systems, in conjunction with computerized physician order entry tools, have been show to decrease medical errors, improve efficiency, and enhance clinical performance. However, CPOE systems can also cause a number of unintended adverse consequences. Although such systems can lead to more efficient and safer care, health care is filled with complexity, variations, and exceptions that are not easily captured by idealized processes. Information systems that are too rigid to support such deviations can lead to decreases in quality, along with caregiver resistance and creative workarounds, that together lower the adoption rate and decrease the positive effects of technology.

We hypothesize that many of the unintended consequences are due to a mismatch between system and task flexibility. Similar problems in other industries have led to the concept of Systematic Yet Flexible (SYF) systems, in which the system supports and encourages a systematic approach, while simultaneously allowing for considerable flexibility. Building upon the general design goals for SYF systems, we designed a framework, called SYFSA (Systematic Yet Flexible System Analysis) for analyzing and designing SYF systems by considering the trade-off between systematicity and flexibility.

2.40pm Break
Session 3: Safety and Evaluation
3pm Forget Statistical Significance: Visualization Tools Enhance Drug Safety Surveillance Sheila Weiss -- UMD

Drug safety is a heavily regulated industry. After spending upwards of 15 years and $1 billion to develop a drug from discovery through approval, sponsors must continually monitor and evaluate the drugs evolving safety profile and take appropriate actions. As part of this process, each company must collect and evaluate voluntarily reported adverse events that are potentially related to the drug exposure and submit them to the FDA. The FDA's adverse event reporting database (AERS) is the world's largest repository of such reports (> 5 million reports). Increasingly, data mining is being used on the FDA AERS database to proactively identify drug-event pairs that are reported at greater than expected frequencies. Currently the data mining algorithms in use rely on disproportionality analysis compare the reporting rates for a particular drug-event pair to the reporting rates in the full data set or a select group of drugs. Drug-event pairs which extend above a predetermined threshold are considered "statistically significant" and triaged for clinical consideration. Given the more than 20,000 reaction terms and thousands of drugs, this type of analyses creates a paper trail of many statistically significant pairs that then need to be considered for immediate action, ongoing monitoring, or dismissal. Misinterpretation of these data and data mining results can be costly, leading to regulatory and legal actions. Because of the stakes, there has been significant research (and professional disagreements) on how to measure the sensitivity and specificity of the various data mining algorithms and methods, creation of "gold standards" of true positive drug-event pairs, and studies assessing the cost of processing "false positives." This is a static and restrictive use of data mining, which is best used for knowledge generation and exploration. In this session, we explore how replacing thresholds and statistical significance with clinician-driven data visual exploration can maximize the potential of AERS for discovery and hypothesis generation.

3.20pm Using Process Definition and Analysis Techniques to Reduce Errors and Improve Efficiency in the Delivery of Healthcare Leon J. Osterweil -- Department of Computer Science, University of Massachusetts, Amherst

As has been widely reported in the news lately, heathcare errors are a major cause of death and suffering, and healthcare inefficiencies result in escalating costs. In the University of Massachusetts Medical Safety Project, we are investigating if process definition and analysis technologies can be used to help reduce heathcare errors and improve heathcare efficiency. Specifically, we are modeling healthcare processes using the Little-JIL process definition language and then analyzing these processes using model checking (with our FLAVERS finite state verification tool), fault-tree analysis, discrete event simulation, and other analysis techniques. Working with the UMASS School of Nursing and the Baystate Medical Center, we are undertaking in-depth case studies on error-prone and life-critical healthcare processes. In many ways, these processes are similar to complex, distributed systems with many interacting, concurrent threads and numerous exceptional conditions that must be handled carefully.

This talk describes the technologies we are using, discusses case studies, and presents our observations and findings to date. Although presented in terms of the healthcare domain, the described approach could be applied to human-intensive processes in other domains to provide a technology-driven approach to process improvement.

3.40pm NIST Advances in Measuring, Evaluating and Improving the Usability of Electronic Health Records Matt Tyrone Quinn, Usability Scientist -- National Institute of Science and Technology (NIST)

Usability represents an important yet often overlooked factor impacting the adoption and meaningful use of electronic health record (EHR) systems. Numerous studies have identified deficiencies in both the usability of EHR systems and key shortcomings among certified EHR vendors in the processes, practices and use of standards and best practices with regard to usability and human factors. Without usable systems, healthcare providers and patients cannot gain the full potential benefits of EHR systems and the investments of the HITECH Act.

NIST established the Health IT Usability Initiative to advance measurement science and develop a framework that defines and assesses health IT usability. The initiative has a goal of creating detailed specifications of an objective, repeatable procedure for measuring and evaluating the usability of health IT systems.

In this session, Matt Quinn will provide an overview of NIST's efforts to date, including recently published technical guidance on "Technical Evaluation, Testing and Validation of the Usability of Electronic Health Records" (NISTIR 7804) and upcoming guidance on the design and validation of EHRs for use with children. In addition, Matt will provide describe feedback from the field on NIST technical guidance and its incorporation into proposed rules for EHR certification.

4pm Discussion & Future Steps
4.40pm Closing

This agenda is for Electronic Health Record Informatics Workshop.