Thursday Scientific Sessions
Thursday, October 11, 2012
7:30 am – 8:55 am
View Scientific Session Videos:
Toronto Room: Applied Informatics and Clinical Workflow
Data Models, Bioinformatics and Personalized Medicine
Location: Regency Ballroom A
ImageJS: Personalized, Participated, Pervasive and Reproducible Image Bioinformatics in the Web Browser
Jonas S Almeida, PhD (email@example.com), James R Hackney
Department of Pathology, Division of Informatics, University of Alabama at Birmingham, Birmingham, AL
The ImageJS tool reported in the July 2012 issue of Pathology Informatics (JPathologyInformatics 3:25) describes a successful attempt to use a Web App ecosystem approach to the dissemination/deployment of Pathology Informatics applications. This approach, validated there wi an image analysis application, only became a realistic proposition with the development of the modern web browser combined with the recent emergence of the third generation of Web technologies (Web 3.0). The key advantages of this approach derive from the browser’s reliance on the code migrating to the machine where the access to the data already exists. This decreases the exposure of sensitive data, which no longer leaves the protected clinical informatics environment. It also facilitates having multiple partners, for example, computational statisticians, develop individual modules which the user can then compose as needed for specific applications.
The basic module of ImageJS, which can be obtained by directing the browser toimagejs.org, includes the module orchestration methods that will respond to the URL of additional modules. This illustration can also be visualized in the two webcasts listed in that web page where multiple combinations of modules, for example to calculate cellular proliferation in KI67 labeled images, are assembled into a single link. As a result, a pathologist would only have to direct the browser to a tailored link to assemble an advanced image analysis application configured for a specific problem: with no download or installation. This gives substance to the argument that Web App ecosystems, such as ImageJS, have significant advantages over more conventional approaches, such as NIH’s ImageJ, as a vehicle to deploy Pathology Informatics applications.
This presentation describes and expands on the ImageJS tool reported in the July 2012 issue of Pathology Informatics (J Pathology Informatics 3:25). Modern web browsers now include code interpreters and support code distribution architectures that are advantageous as a computational environment to deliver informatics applications. In a nutshell, this presentation explores the use of web Apps to deliver image analysis directly (no downloads or installations) to where it is needed. We will argue, with examples such as determining proliferation from KI67 labeling images, that this approach has important advantages over conventional systems such as NIH’s ImageJ.
A core module is delivered with the imagejs.org URL with analytical and user interface components being then loaded with their own script tags. The effect is that a particular image analysis solution is automatically configured from a string of URL’s, blurring the distinction between a “personalized” application and an analytical workflow.
A configurable Web App ecosystem for image analysis, inviting module development by statisticians and human-machine interface researchers alike, is made available with open source and in the public domain. This result is detailed in a July 2012 Pathology Informatics report, and is also illustrated by a number of YouTube webcasts, including the two listed in http://imagejs.org. This ability of non disruptively delivering advanced computational statistics applications to the point of care also caught the attention of the media which described it as the “angry birds” approach.
The emergence of the modern web browser coupled with the more open architectures of the third generation of Web technologies creates novel opportunities for pathology informatics. The ImageJS application to image analysis suggests that ecosystems of Web Apps have fundamental advantages as informatics solutions as concerns deployability, configurability, reproducibility and protection of sensitive personal health information (PHI).
ClusterFASTQ: A Method for the Identification of Translocations in Clinical Next Generation Sequencing Data
Eric J. Duncavage, MD (firstname.lastname@example.org)1, Haley Abel, PhD2
1Department of Anatomic and Molecular Pathology,
Washington University College of Medicine,
St. Louis, MO
2Division of Statistical Genetics, Washington University
College of Medicine, St. Louis, MO
The identification of gene translocations has both diagnostic and prognostic significance in molecular oncology. Current clinical methods for translocation detection generally rely on low-resolution techniques such as fluoresence in-situ hybridization (FISH), however recent work has demonstrated that translocations can be identified by targeted next generation sequencing (NGS). Existing methods for translocation detection in NGS data involve the identification of discordant paired-end sequencing reads, however, these methods are subject to a high false positive rate that makes them unsuitable for clinical molecular diagnostics. ClusterFASTQ improves upon existing methods and offers a greatly reduced false positive rate while maintaining a high sensitivity.
ClusterFASTQ is a command line utility implemented in Java. It accepts aligned sequence (BAM) files as input and outputs the chromosomal positions of both translocation partners as well as a contig spanning the breakpoint sequence.
Many existing methods for the detection of translocations in next generation sequencing data rely on the identification of discordant paired-end reads where paired reads map to different chromosomal regions. While such methods are sensitive and computationally efficient, they produce a high number of false positive results due to repetitive elements in the human genome. Other methods reduce the false-positive rate by looking for chimeric single-end reads, but the performance can be heavily-dependent on the choice of aligner, and the methods generally do not produce large breakpoint-spanning contigs. ClusterFASTQ improves upon these methods by first identifying clusters of discordant paired reads, then validating the clusters by re-mapping the unmapped or incompletely-mapped partners of nearby reads. Finally, it assembles all reads mapping to the vicinity of the cluster into a contig, typically 300-3000 bp long.
DNA sequences from 8 formalin-fixed cancers with known ALK or MLL translocations was analyzed by both Breakdancer and ClusterFASTQ. Both tools correctly identified the translocations, however Breakdancer produced an average of ~1200 putative translocations/case while ClusterFASTQ identified only one/case.
ClusterFASTQ allows for the rapid detection of translocations from targeted next generation sequencing data. Using a secondary single-end read re-alignment step, ClusterFASTQ greatly reduces the number of false positive translocation results. obtained, thereby permitting NGS-based translocation detection in the clinical laboratory.
Patients Accessing Laboratory Results via Patient Portal: What are the Risks?
Maggie L. Hopkins, MD, MBA (email@example.com)1, Alexis B. Carter, MD2
1Department of Pathology, University of Utah School
of Medicine, Salt Lake City, UT
2Department of Pathology and Laboratory Medicine,
Emory University School of Medicine, Atlanta, GA
Patient portals (PTPRs) are web portals typically offered by a healthcare entity to allow patients to view their entity-related medical information. The use of PTPRs is increasing, in part because of programs established by the U.S. government including Meaningful Use (MU) of Electronic Health Records. Several objectives of MU Stage 1 and recently finalized MU Stage 2 require that a percent of patients have access to a PTPR that includes lab results. However, controversy exists on whether direct release of medical information exposes patients to risks of misinterpretation or other harm. To answer this question, the authors undertook a review of the existing medical literature.
PubMed (http://www.ncbi.nlm.nih.gov/pubmed/) and SurveyMonkey (http://www.surveymonkey.com) were used to review the literature and collect data, respectively.
Recent medical literature on PTPRs was retrieved from PubMed by searching for terms in quotes: “patient portal” and “patient portals.” A total of 98 articles were retrieved and filtered down to 34 based on the whether the title was relevant to the topic. Subsequently, two independent reviewers reviewed each of the 34 articles. Articles were excluded if their focus was on manual entry of data by the patient (e.g., personal health record), electronic visits or if the full text could not be acquired. After exclusion, 10 articles remained which were categorized according to their focus.
The papers were reviewed for whether or not they covered the following items, and the percent which did address was recorded: Lab results available: 50%(5), Ability to access portal: 70%(7), Described patient demographics: 50%(5), Reported access disparities: 40%(4), Described usability: 60%(6), Access to sensitive results: 20%(2), Compliance with laws or standards: 10%(1), Specific vendor platforms: 10%(1), See Table 1
Peer-reviewed literature indicates that patients are most likely to use PTPLs to view laboratory results. Despite this, peer-reviewed articles examining how patients interpret and use laboratory results are lacking. Limitations of this study include lack of standard terminology and overlap with “personal health record(s)”, thereby likely artificially reducing the number of articles for review. Regardless, much more work is needed to determine compliant standards by which laboratory data should be displayed in PTPLs for ease of patient comprehension.
Development of Image Analysis for Epidermal Growth Factor Receptor Expression as a Potential Predictive Biomarker in Colorectal Cancer
Ryan A.J Hutchinson, Bsc (firstname.lastname@example.org)1, Yinhai Wang, PhD1, Jacqueline A James, PhD1, Manuel Salto-Tellez, MD1, Richard A. Adams, MD2, B. Jasani, MD2 and Peter W. Hamilton, PhD1
1Centre for Cancer Research and Cell Biology, School of Medicine, Dentistry and Biomedical Science, Queen's University Belfast, Belfast, United Kingdom
2Institute of Cancer and Genetics, Section of Oncology and Palliative Medicine, Cardiff University School of Medicine, Cardiff, United Kingdom
The COIN trial was a major multinational study examining the comparison of continuous chemotherapy with and without Cetuximab to intermittent chemotherapy as a first line therapy in previously untreated advanced colorectal cancer patients. Original findings from the COIN trial suggested that pathology-based visual scoring of Epidermal Growth Factor Receptor (EGFR) expression using immunohistochemistry (IHC) in advanced colorectal cancer tissue microarrays, did not indicate any predictive value for treatment with cetuximab and chemotherapy in first line therapy. The aim of our study was to examine whether automated IHC image analysis has the potential to provide new evidence as to the utility of EGFR IHC as a predictive biomarker.
The complete set (27 slides) of IHC stained, advanced colorectal cancer tissue microarray slides were digitised using an Aperio ScanScope CS2 scanner. Representative regions were randomly marked up across 23 tissue microarray cores by the original COIN trial pathologist. A total of 70 regions were annotated which were representative of a range of staining intensities. The digitised IHC stained slides were imported into Definiens Tissue Studio for EGFR quantification. This cohort of marked up tissue microarray cores were used as a training set to develop the algorithm. A custom region recognition method applied to EGFR IHC images identified histological regions and a novel image analysis solution was created to identify positive membrane staining using subsets of identified regions within the EGFR IHC images.
The defined algorithm was capable of accurately segmenting and measuring positive cell membrane EGFR expression across the range of colorectal cancer samples. This allowed for the successful selection of epithelial tumour cells and exclusion of positivity in stromal/non tumour regions. A direct comparison between visual IHC scores and computerised image analysis derived IHC scores showed a strong correlation (r=0.96, P<0.0001). Some results suggested that visually marked up cores with a lower histological score were scored higher in some instances using image analysis.
The application of image analysis facilitates the automated quantification of the EGFR IHC expression, providing accurate membrane segmentation and measurement. Although visual and automated IHC evaluation showed strong concordance in this sample set, the clear advantage of image analysis is reproducibility and consistency across large numbers of samples. Having now established a reliable image analysis method for quantification of EGFR expression in colorectal cancer, future work will examine the role of EGFR image analysis in predicting response to cetuximab in the COIN trial patient cohort.
A DICOM Prototype in XML with Relationships
Robert C. Leif, PhD (email@example.com), Stephanie H. Leif, MS
Newport Instruments, San Diego, CA
CytometryML consists of XML schemas that describe the specimen processing, acquisition, and results of cytometric measurements. Presently these measurements can be performed with a flow cytometer or digital microscope. Because of the considerable overlap between cytometry and digital microscopy, data-types have been incorporated from both the International Society for advancement of Cytometry, ISAC, and the Digital Imaging and Communication in Medicine, DICOM, standards. The DICOM separation of studies and instances has been included. The architecture of CytometryML now includes the capacity to describe relationships in a manner, which is similar to that of the Resource Description Framework (RDF) and the ZIP file container for both the meta- and binary data is based on the popular EPUB standard.
The schemas are written in the XML Schema Definition (XSD1.1) language and validated to demonstrate adherence to XSD1.1. Their content was tested by translating specific XSD elements into XML and filling in the values of the objects contained therein. The use of an element based implementation of relationships permits bidirectional and multiple relationships between two objects to be expressed.
Modularity of the design was enhanced by basing the schemas on objects. The DICOM hierarchical separation of a metadata containing series and separate instances, which also include binary data was maintained. Because of compatibility with draft Supplement 161, CytometryML should be able to read XML data that is stored on the picture archiving system, PACS while minimally loading the server, as well as being compatible with future efforts to write XML data on the PACS.
An XML based system that incorporates data-types from existing standards and provides enhanced, but simple to understand relationships has been created. Very preliminary data indicates that these XML data-types can be used with XHTML5, which would permit the creation of a medical informatics system that has access to the full power of the Internet.
CytometryML can be considered as a basis for a collaborative effort between pathologists and cytometrists for development of a continuum of complimentary interoperable standards and a prototype of a future Internet based version of DICOM, DICOM 4.
Best Block Designation in Surgical Pathology: a Help or a Hindrance to Subsequent Molecular Studies?
Andrew M. Quinn, MD (firstname.lastname@example.org), Frank Kuo, MD, PhD
Department of Pathology, Brigham and Women's Hospital, Boston, MA
Selecting blocks with tumor suitable for testing is a time-consuming step in molecular diagnostics workflow. In order to improve turnaround time, our department instituted a quality improvement initiative in 2010 by which surgical pathologists select a "best block" (BB) for molecular testing and record its tumor content during case finalization. The purpose of this study was to compare pre-analytical and analytical processing intervals between molecular cases for which a BB was designated and those for which it was not.
Two partially-automated, real-time tracking systems were employed: an in-house, web-based laboratory information management system for following pre-analytical and analytical molecular case processing time points, and an in-house enhancement to Sunquest PowerPath (Sunquest Information Systems, Tucson, AZ, USA) for BB designation and characterization. Time points were logged via electronic verification within the web-based laboratory information management system. Data analyses were performed via GraphPad Prism (GraphPad Software, Inc., San Diego, CA, USA).
Retrospective review of time points for 2,185 consecutive tumor genotyping cases (August, 2011 to July, 2012) at a large academic medical center (Brigham and Women's Hospital, Boston, MA, USA) was performed. Time points include: report review, block/slide adequacy review, pathologist approval for molecular analysis, molecular analysis start and end (DNA extraction and sequencing), and molecular report finalization. Cases were divided into three groups: BB designated and ultimately used (BB Used), BB designated but not used (BB Not Used) and BB not designated (No BB). Non-parametric analysis of variance calculations with post-hoc Dunn's multiple comparison tests were used for comparison with significant p values of less than 0.05.
A summary of the results is presented in Table 1.
"Best block" designation at the time of surgical pathology case finalization results in significant improvement in the overall turnaround time for subsequent molecular analyses. The improvement is best realized in the pre-analytical phase of reviewing a case, but may reflect selection bias. Surprisingly, BB Not Used cases have significantly shorter pre-analytical intervals than BB Used cases, suggesting a confounding variable. The variability of numerous pre-analytical intervals (data not shown) indicates that further refinements and subsequent review of the tracking process are needed for optimization.
Applied Informatics and Clinical WorkflowLocation: Acapulco Room
How Can Laboratories Help Their Hospital Decrease Readmission Rates for Heart
Eugenio Zabaleta (email@example.com),
John Burgess, MD, Kay Truax, MS and Michael Patterson, DO
MedCentral Health System, Mansfield, OH
Starting third quarter of 2012 hospitals will experience decreased Medicare reimbursement for all Medicare discharges when their readmission rates are higher than expected according the Hospital Readmissions Reduction Program. In response MedCentral Health system developed Heart Success, which is a program to assist patients with heart failure (HF) to avoid readmission to the hospital. As well, MedCentral utilizes IT to help identify all HF patients that could benefit with this program.
To respond to the new health care challenges, MedCentral has decided to use a Laboratory Decision Support Software called RippleDown® (from Pacific Knowledge Systems) to help identify all HF patients that could benefit with this program. Within this software, rules were developed for the purpose of stratifying patients according to their BNP and Troponin results (Circulation 2003; 108; 833-838). Also, this software will automatically communicate this information to the Heart Success personnel with minimal obstruction to clinical operation.
Since mid October 2011, MedCentral Laboratory is able to perform risk stratification on every inpatient that has a BNP test done (BNP is the trigger). Once the risk is calculated the result will be sent via e-mail with patient demographic information, risk calculation, and other laboratory data to the Heart Success personnel to prioritize patient management.
The US national Heart Failure readmission rate is 24.8 % and can be as high as 40 % according to CMS (data from 2011). MedCentral hospital rate for people not enrolled in the Heart Success project is 26.7 %, while it is 4% for the enrolled patients. This project went live on October 2011 and since then the hospital average of HF readmission rate is 20% or below (20.0% in October, 13.3% in November, and 16.1% in December). No data is available for 2012.
Even though most of the patients recruited for Heart Success are through physician referrals, the automated CHF patient risk stratification project serves as a safety net for identifying those patients recently diagnosed with CHF or those who are not referred, or for those not treated in our main hospital.
Using Camera Systems to Read Entire Batches of Barcodes & More
Lyman T. Garniss, BA (LGarniss@partners.org)
Partners Healthcare Systems – IT, Massachusetts General Hospital, Boston, MA
Massachusetts General Hospital (MGH) Pathology Service is now using 2-D barcodes on all cassettes and slides. We are also tracking and routing these materials using Sunquest CoPath SMART. In some areas it is desirable for patient safety and for workflow efficiency to scan entire batches of cassettes and/or slides instead of scanning individual assets.
The technology to accomplish this project included; A Cognex industrial camera system with software that enables the identification and reading of 2-D barcodes. The Sunquest CoPath SMART system, unique asset identifiers for specimen cassettes and slides encoded on 2-D barcodes; Boston Workstation scripting tools, and an enclosure created by Machine Vision Consulting
A rack of barcodes cassettes or a folder of slides are loaded into the camera enclosure. The rack or folder activates a switch in the enclosure, the system is activated turning on the lights and the picture is taken. The picture is transferred to a computer, analysis is performed on the picture and the barcodes and a list of asset numbers is produced and sent to a file.
The file is then read by a Boston WorkStation script and then checked against the Sunquest CoPath database to ensure all specimens are together (none are missing) and the script also " checks" the assets out of the area in the tracking system. The system can complete the entire process of reading, and checking and updating the CoPath database on 78 cassettes in 2 minutes. Doing this manually takes a
human over 10 minutes and does not allow for ensuring that all cassettes that should be together are together.
Batching reading of Pathology assets (cassettes and slides) using an industrial camera system increases patient safety and allows for faster through put of specimen processing when using tracking and routing software.
Clinical Genomicist Workstation
Rakesh Nagarajan, MD, PhD (firstname.lastname@example.org)1, Mukesh K. Sharma, PhD1, Joshua Phillips, BS2, Saurabh Agarwal, BTech2, Wesley S. Wiggins, MS2, Savita Shrivastava, MS1, Sunita B. Koul, MS, BS1, Madhurima Bhattacharjee, MS1,
Caerie D. Houchins1, Raghavendra R. Kalakota, MS1, Bijoy George, MBA, BS1, Rekha R. Meyer, PhD1, David H. Spencer,MD, PhD1, Christina M. Lockwood, PhD1, TuDung T. Nguyen, MD, PhD1, Eric J. Duncavage, MD1, Hussam Al-Kateb, MS, PhD1, Catherine E. Cottrell, PhD1, Suhasini Godala, MS2, Ravi T. Lokineni, MS2, Sameer M. Sawant, MS2, Vasudev Chatti, BTech2, Suresh Surampudi, PhD2, Raja Rao Sunkishala, MS2, Ramakant Darbha, MS2, Sharath Macharla, BTech2, Jeffrey D. Milbrandt, MD, PhD3, Herbert W. Virgin, MD, PhD1, Robi D. Mitra, PhD3, Richard D. Head, MS3, Shashikant Kulkarni, MS, PhD, FACMG1, Andrew Bredemeyer, PhD1, John D. Pfeifer, MD, PhD1, Karen Seibert, PhD1
1Department of Pathology and Immunology,
Washington University in St. Louis, St. Louis, MO
2SemanticBits LLC, Herndon, VA
3Department of Genetics, Washington University in
St. Louis, St. Louis, MO
The use of NextGen Sequencing clinically necessitates the need for informatics tools that support the complete workflow from sample accessioning to data analysis and reporting. To address this need we have developed Clinical Genomicist Workstation (CGW).
Apache 2, Tomcat 6, Grails 2.0, Spring 3.1, Drools 5.2, YUI 2, jQuery 1.7, HTML5, Groovy 1.8, FreeMarker 2.3, Java 6, Oracle Database 11g, Red Hat Enterprise MRG 7.6.3, Integrated Genomics Viewer 2.1
CGW is a secure, n-tiered application. A web browser submits requests to application servers that persist data in a relational database. CGW utilizes Grail and Groovy, permitting advanced, automated, and customized access to the application’s features. RedHat MRG is used to launch and track complex analysis pipelines.
CGW is used by Washington University Genomic and Pathology Services for clinical genomic testing of many cancers. Orders are accessioned in CoPath and subsequently, the CGW. DNA from specimens is subjected to NextGen sequencing, which is tracked in the CGW. Informatics pipelines for variant detection are executed from the CGW once the sequencing run completes. Results are loaded into the CGW database following analysis completion. Called variants are viewed in context of data from other variant databases such as dbSNP, COSMIC, HGMD, and ClinVar, and previously curated interpretations for variants are automatically inserted into a draft clinical report. The draft report is reviewed and finalized by the clinical genomicist and messaged to the medical record for persistence. CGW has been used to sign out over 100 cases since November, 2011 and more than 2000 cases are expected to be signed off by June 2013. There are 22 ordering oncologists and 7 clinical genomicists that use the CGW. Future plans include increasing the number of comprehensive cancer panel genes from 27 to 50 and introducing test panels for drug metabolism and transport, renal and cardiac diseases. Finally, CGW is in the process of being adopted at several other clinical genomic laboratories nationally and globally.
CGW is a “soup-to-nuts” software solution for managing clinical genomic testing through order intake, sequencing, analysis, case review, interpretation, and reporting.
Development of Automated Testing Solution for Anatomic Pathology Laboratory Information System (AP-LIS): Background, Technical Design, and Future Vision
J. Mark Tuthill, MD (email@example.com)1, Ron Brown1, Jennifer Lyle2, Chris Gardenhour
1Division of Pathology Informatics, Department of
Pathology and Laboratory Medicine, Henry Ford
Hospital, Detroit, MI
2Software Testing Solutions, Tucson, AZ
We have previously reported on use of an automated testing solution for clinical pathology CLIA validation testing. Working with our vendor partners we developed a novel testing solution to support validation testing of our CoPathPlus LIS, specifically for the surgical pathology cases. While validation testing is less complex for AP-LIS, functional testing is becoming increasingly complicated as AP-LIS increase in sophistication.
AP-LIS: Sunquest CoPathPlus 6.0 (Sunquest Information Systems, Tucson, AZ); STS Workstation, 1 Dell Optiplex 745 (Dell, Roundrock, TX) running Microsoft Windows XP; STS™ Lab Volume and Scenario Solutions (Software Testing Solutions™, Tucson, AZ); Remote Connection: PC Anywhere v11.5 (Symantec™, Cupertino, CA).
Working with STS their software was adapted to validate aspects of the CoPathPlus 6.0 AP-LIS. Specifically the system will automatically accession cases across all specimen classes inserting all part types and quick text in that specimen class. Further, the system will automatically and iteratively order all available stains, stain protocols and procedures using a specified number of accessions. If a stain has been previously used in a protocol, it will skip to the next stain in the dictionary. All cases are automatically signed out and transmitted across the HL7 interface. The design leverages all rules system configurations built in the AP-LIS. For examples a quick text associated with a part type will be used preferentially to random quick text. The system can be uniquely configured to test workflow at different sites.
While the system is very useful for validation of results in a receiving system by generating a large amount outbound data, a more important benefit is assessment of functionality of AP-LIS. Since the system leverages rules within the AP-LIS, secondary events are generated that can be difficult to test in a manual fashion. All of the cases signed out automatically in STS also generate events such as billing data, slide label and cassette printing, and transactions to interfaced systems such as auto-stainers. Manually, this would take weeks to actualize; as automated in STS, we can perform the same testing for every part type, stain protocol and procedure with all secondary events in less than 48 hours with using the least possible number of accession numbers.
Automation of AP-LIS validation testing can provide significant benefits both in terms validation testing as well as system functional testing. The solution we continue to be developed and enhanced to all it to be applied cytopathology testing.
APPLIED TOPICS SUBTRACK:
Applied Informatics and Clinical Workflow
Location: Toronto Room
Mitigating the Risks of Custom Software Development in Pathology using Open Source Software
Peter Gershkovich M.D (firstname.lastname@example.org), John Sinard M.D., PhD
Department of Pathology, Yale Medical School, Yale University School of Medicine, New Haven CT
The demands of patient care, financial constraints, and educational challenges produce an increasing pressure to create better technology to handle multiple missions of Pathology Departments. Over the past eight years, the Pathology Informatics group at Yale has developed many applications to handle Outreach Support, Barcoding and Tracking of blocks and slides, Frozen Section Management, Residency Training tools, Document Management, Monitoring and Notification, etc. This development relied heavily on Open Source software. Using well-tested, production level and enterprise quality common components not only increased the productivity of development but also reduced the amount of custom code. We argue that it in turn mitigates the risk of building and supporting custom software.
The overwhelming majority of our custom applications and modules have been developed using Java. Additional open source applications, APIs, and Frameworks included: Apache HTTPD Server, Apache Tomcat, MySQL database, Quartz scheduling framework, Google Web Toolkit, HAPI API, iText, jFreeChart, Spring, Hibernate, etc.
Most of our applications have a user interface build with Google Web Toolkit. The code is typically deployed on Apache Tomcat following the Servlet 3.0 specification. The middle tier comprises business logic and uses various APIs for common tasks (e.g. iText for PDF rendering, jFreeChart for creation of reports, etc.). An application can have its own Relational Database Management System and/or have access to our core AP LIS database deployed on Sybase ASE 15.03. All custom databases are implemented on MySQL 5.5.
None of the project modules took longer than six months from the design to the implementation. Custom development allowed implementation of small functional blocks of code that encapsulated a particular workflow step creating immediate improvements in the process and making it easy to reengineer the process and implement the entire solution.
The challenges of Pathology Departments at Academic Institutions require development of Information Technology Solutions. Build vs. Buy approaches have pros and cons. The evolution of Open Source software creates an opportunity for rapid development unparalleled by either of the extremes and reduces the risks of in-house application development.
Implementation of Customer Relationship Management (CRM) Software to Support Laboratory Outreach at Henry Ford Health System (HFHS), Henry Ford Medical Laboratories (HFML)
Mehrvash Haghighi, MD (email@example.com),
J. Mark Tuthill, MD, Sarah Mackay
Department of Pathology and Laboratory Medicine, Henry Ford Hospital, Detroit, MI
To support our laboratory outreach efforts, we implemented and customized Microsoft Dynamics CRM application to automate and standardize the process of capturing and resolving defects and client issues. This application provides: issue tracking; management and escalation; customer data management, customer configuration data, and sales force support including marketing, call tracking and request management.
Microsoft Dynamic CRM 3.0 (3.0.5300.0) purchased through Sunquest Information Systems, Tucson, AZ; Screens and workflow customized by HFHS; Deployed on Windows VMware servers (DL 380 G6, (2) 6-core processors, 24GBRAM, (8) 146 GB drives.)
We modified the design of CRM based on HFML needs and workflow. We engaged users providing opportunities for decision making, key requirements, current headaches, and preferred work flow for incorporation into our final design. Communications and training strategies were developed to disperse information, and guide users for successful adoption including e-mail announcements, demonstrations, training classes, and quick guides. Current workflow and issues and problems were analyzed by direct observation and parallel testing to allow for effective design.
The implemented CRM module has replaced manual processes that included paper logs, spreadsheets, and manual generation of issues trends data. This consolidated several sources of disparate information from Excel, Word and MS Acces programs into a seamless process. As an integrated, consolidated system, it offers clear visibility and management of customer data.
Moving from a paper-based system to an electronic system has enabled HFML client service to strengthen customer relationships and refine its business processes and work-flow to support greater profitability. Key benefits can be divided into direct benefits such as reduced paper, printing and storage costs and indirect benefits including improved customer service productivity, quality assurance, customer retention and relationships. CRM has created seamless connectivity and communications between clients, customer service, couriers and lab personnel through a paperless process automating activities regarding client issues. It also offers a centralized location for sales and marketing notes. By using CRM employees are better informed regarding the client data and issue history.
The Department of Health and Human Services “Wall of Shame”: An Analysis of Large Security Breaches of Protected Health Information
Chiraag D. Patel, MD (firstname.lastname@example.org)1, Alexis B. Carter, MD2
1Department of Pathology, Stony Brook University
Medical Center, Stony Brook, NY
2Department of Pathology, Emory University School of
Medicine, Atlanta, GA
Health data contains personal information that could be used to harm a patient. Concerns over the ever increasing use of electronic media to manage health information resulted in several federal laws, one of which (Health Information Technology for Economic and Clinical Health Act (HITECH)) requires notification of patients if their health data has been disclosed in an unauthorized manner (security breach). If a security breach greater than 500 patients, details are made publicly available on the U.S. Department of Health and Human Services (HHS) website. An analysis of this public data set has not been previously published.
The dataset of all security breaches involving 500 or more patients from inception of HITECH in 2009 was downloaded in March 2012 from the U.S. Department of HHS website. These breaches were analyzed and categorized by year as well as by the type of loss, medium of loss, and number of affected patients.
Over 20,066,249 patients’ data were breached, but it was not possible to determine how many of these represented unique patients (maximum of 6.4% of U.S. population affected). Loss of large sets of patient data was primarily due to theft (52.18%) or accidental disclosure (18.39%). The location of the data at the time of loss was overwhelmingly on paper (25.29%) or on a laptop (25.06%). Further detail is present in Table 1. The number of reported large security breaches has increased each year, and the most significant increase in overall security breaches occurred from 2009 to 2010 (252.13%).
Table 1. Number of patients affected by security breaches by type and location of data.
Recently enacted federal regulations have given the health care industry a much needed push into electronic media to facilitate improvements in patient quality and safety. However, patient harm may occur if sensitive data is accessed or distributed in an unauthorized manner, and health care entities which must disclose security breaches to patients will incur financial, legal and reputational difficulties. The analysis of this data set reinforces the security risks inherent with maintaining sensitive data in any format especially data on portable electronic devices. Limiting unwanted security breaches and unauthorized data access should be a high priority for all health care entities.
Automating Grossing in Anatomic Pathology
Shree Sharma, MD (email@example.com)1, Manisha Singh, MBBS2
1Department of Pathology, University of Arkansas for
Medical Sciences, Little Rock, AR
2Internal Medicine, University of Arkansas for Medical
Sciences, Little Rock, AR
More than 50% of the specimen load of an anatomic pathology laboratory is constituted by biopsy specimens. The present proposal is related to automation of the anatomic pathology laboratory. The automation will handle the grossing of small specimens not requiring orientation and inking. The current problems in grossing include shortage of staff, mix-up of specimens, and the time devoted to handling of specimens.
The technology includes redesigning the existing devices which are used in the laboratory on the day to day basis.
The design includes a biopsy test tube, cassette and the robotic system to transfer the specimens. The biopsy- test tube will have an in-built biopsy bag. The biopsy will be collected by the surgeon in this biopsy-test tube with patient identifiers and barcode. The biopsy-test tube after getting accessioned in the pathology lab will be put on the conveyor belts, similar to that used in clinical pathology. A second row of belts with cassettes will move along on a separate belt. The specimen bag will be pulled out of the test tube and dropped in the labeled cassette automatically by the robotic system.
The proposed design and technology will decrease the time needed for processing specimens and will improve the turnaround time. It will eliminate the chances of specimen mix-up and carry over because the specimen will not be opened on the grossing table. It will also reduce the problems related to shortage of staff in the grossing laboratory.
The rapid advancement in the field of digital pathology will change the way we practice pathology in future. The automation of grossing in anatomic pathology is much needed and the lack of attention towards this aspect will act as a bottle neck in the automation and development in the field of anatomic pathology.
Customization of a Digital Pathology Consultation Portal for Clients and Consultants
Somak Roy, MD1 (firstname.lastname@example.org), William Cable, BS2, Eugene Tseytlin, MS3, Andrew Lesniak, BS2, Jeffery McHugh2 , Gonzalo Romero Lauro, MS, MBA2, Samuel Yousem , MD1, George Michalopoulos, MD, PhD1, Anil V. Parwani, MD, PhD1, Liron Pantanowitz, MD1
1Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh, PA
2Information Services Division, University of Pittsburgh Medical Center, Pittsburgh, PA
3Department of Biomedical Informatics, University of Pittsburgh, Pittsburgh, PA
Peer review and obtaining second opinions on perplexing cases is an integral part of contemporary anatomical pathology practice. The era of digital pathology facilitates inter-institutional and international pathology teleconsultations. However, a successful telepathology program requires significant customizations to be effective. This study reports our experience with a highly customized web-based portal for handling digital second opinion consultations.
The portal (https://kingmed.upmc.com) is a web-based interface coded in ColdFusion which enables telepathology by incorporating clinical workflow with a whole slide image (WSI) viewer. The solution leverages a vendor neutral viewer that is compatible with the most common imaging pathology formats available in the market. This is an internally-developed Java applet which supports navigation, annotation and snapshot functionality in addition to advanced features such as the latest NDPI file format from Hamamatsu. To improve performance, multi-threading capability was added to the viewer.
The portal is based on a client-server model. Using secure login, the client (Kingmed Laboratories, China) scans the slides to an image server and accessions their case with accompanying clinical data into the web-based system. Pathology consultants (UPMC, Pittsburgh) access the WSI on the image server via the web-based collaboration tool and employ the portal to manage workflow (e.g. triage requests), information exchange (e.g. request more history, slides or stains), navigate images (e.g. annotation, snapshot acquisition), for reporting, and transmission of diagnoses to the submitting client.
A total of 33 cases were received for consultation via the portal over 8 months. All cases were surgical pathology specimens, most commonly breast/female genital tract (n=10, 30%) and soft tissue (n=6, 18%). WSI (n=365, average 11 slides/case) received included 134 H&E, 6 histochemical and 225 immunohistochemistry slides. Mean turnaround time for 22 consulting pathologists was 40 hours (range 2-152 hours). Delayed cases were due to network outage problems. Post-launch feedback resulted in further customization to incorporate transcription services, peer-to-peer review for consultants, provision for issuing addenda/amendments and improved WSI viewing experience.
Combining digital imaging with the Internet to create a custom client digital pathology portal broadened our institutions capability to offer international teleconsultation services. This scalable, secure, and user-friendly web-based collaboration tool permitted cases to be handled via telepathology in a vendor agnostic manner. Ongoing front- and back-end customization of this solution is responsible for client satisfaction and consultant buy-in.
Automated Review of Laboratory Test Appropriateness
Ronald Hauser, BSE, MD (email@example.com)
Department of Laboratory Medicine, Yale-New Haven Hospital, New Haven, CT
Inappropriate use of laboratory services produces billions of dollars in medical waste annually. The determination of the appropriateness of a laboratory request involves a cost-benefit analysis placing relative utility on the test cost, iatrogenic injury in specimen procurement, the potential for misleading results, and favorable changes to patient management prompted by test outcomes. Weighing these factors results in a threshold that varies with pre-test probability. In everyday practice, clinicians use the patient presentation to assess the pre-test probability and recommend lab tests. Quantification of this intuitive process forms a barrier to automation. This paper proposes an alternative approach using Bayes’ Law to retrospectively calculate pre-test probability using test outcomes.
The application of this approach to laboratory information systems could systematically identify low-yield tests for utilization review.
The study compared the pre-test probability of West Nile meningoencephalitis in a patient population referred for West Nile IgM from cerebrospinal fluid using clinical features and Bayes’ Law. A literature review identified clinical signs with high sensitivity for West Nile meningoencephalitis: seasonality (August, September), age (≥50), fever (≥100.4F before or during evaluation), and CSF pleocytosis (≥5/mm3). A retrospective chart review of 439 test requests recorded these signs. To obtain the inputs to the Bayes’ calculation, test results were obtained. The product insert provided estimates of sensitivity (95%) and specificity (99%).
5 of the 439 tests had a positive result. The Bayesian pre-test probability had a value <1% [(0.99+5/439-1)/(0.99+0.95-1)]. The clinical assessment of pre-test probability found 11 of 439 patients (2.5%) meeting all four criteria for West Nile meningoencephalitis. Of these 11 cases, 4 had a positive result (36%). The single case not satisfying all four features missed only the age criteria by one year.
Pre-test probability calculated with Bayes’ Law and a clinical assessment both predicted a low disease prevalence in the population referred for testing. The application of Bayes’ Laws allows laboratorians to determine the threshold used by clinicians to place requests on laboratory services. It permits the systematic identification of low-yield diagnostic testing, which should prompt a utilization review. Automated determination of pre-test probability increases its feasibility as a pay-for-performance
DECISION SUPPORT SUBTRACK:
Applied Informatics andf Clinical Workflow
Location: New Orleans Room
Real-Time Detection and Display of Critical Values in the Laboratory Improves Caregiver Notifications
Kavous Roumina, PhD (firstname.lastname@example.org); Eugene Farber; Walter H. Henricks, MD
Center for Pathology Informatics, Cleveland Clinic, Cleveland, OH
Timely notification to caregivers of critical test results requiring immediate attention is an important function of the laboratory and necessary for accreditation compliance. In high volume laboratories, improvements in alerting laboratory staff to the presence of unreported critical values could improve time to communication of such results and reduce call back failures. We developed an automated system that detects and displays to laboratory staff pending critical results not yet communicated to caregivers.
Data analysis and presentation components (.NET Framework 3.5, Internet Information Services, Microsoft); relational database (MS SQL Server 2005, Microsoft); laboratory information system (LIS) (Sunquest Laboratory).
The system consists of Controller and Presentation modules, both of which use the same back-end database. The Controller continually queries the LIS, storing in the database only results not denoted as “called.” The web-based Presentation module displays on large monitors color-coded critical results, yellow when first displayed, red within 30 minutes of availability from the analyzer. After caregiver notification, a technologist adds a predefined result modifier in the LIS allowing for the result removal from the display. Easily configurable system parameters enable inclusion of additional test types.
The system processed over 3.5 million results from the LIS (potassium and glucose; Feb. – Jul., 2012). On average, 835 results/hour were analyzed (24 hour cycle), of which 1.7 results/hour were displayed as
critical results. For outpatient critical results, where timely communication is more difficult than for inpatient, the laboratory met monthly targets for percentage results called back (92% within 30 minutes) in 5 of 6 months (83%) following implementation, compared with only 3 of 6 months (50%) in the same months in 2011. For inpatient critical results, the laboratory met targets in all 6 months (100%) compared to 5 of 6 months (83%) in 2011.
We developed an automated system that detects critical results not yet communicated to caregivers and displays them efficiently to laboratory staff through a web browser. Following implementation, the laboratory realized improvements in meeting targets for critical value reporting time. The design paradigm may be extended to “urgent” results as well.
Utilization of Clinical Laboratory Reports with Graphical Elements
Brian H. Shirts, MD, PhD (email@example.com)1; Nichole Larsen, BS2; Brian R. Jackson, MD, MS2,3
1Department of Laboratory Medicine, University of
Washington, Seattle, WA
2ARUP Laboratories, Salt Lake City, UT
3Department of Pathology, University of Utah, Salt
Lake City, UT
Graphical reports that contain charts, images, and tables have potential to convey information more effectively than text-based reports; however, studies have not measured how much clinicians value such features. We sought identify factors that might influence the utilization of reports with graphical elements postulating that this is a surrogate for relative clinical utility of these graphical elements.
ARUP Laboratories established pilot projects to test an online enhanced electronic laboratory reporting system in November 2009 to provide clinicians graphical elements that cannot be transmitted or displayed with current interface systems. A URL along with chart id number and password are provided as a footnote to the LIS-version test result. When a test result is verified within the laboratory information system a pdf-format chart is generated and stored on a server can only be accessed by those that have access to the password transmitted with reports via current interfaces. Login information linked to specific test reports allowed us to track how often these reports were accessed.
We evaluated utilization of enhanced electronic laboratory reports from institutions across the United States. We monitored on-demand clinician access to reports generated for 48 reportable tests over 22 months. We evaluated utilization of 174,170 reports with graphical elements by clinicians at all institutions that use ARUP as a reference laboratory. We used descriptive statistics, regression, and meta-analysis tools to evaluate groups of similar test reports. We compared download rates for tests with different forms of graphical and photographic information to several reports which were generated with standard formatting but no additional clinical information.
Median download rate by test was 8.6% with high heterogeneity in download rates between tests. Test reports with additional graphical elements were not necessarily downloaded more often than reports without these elements. Recently implemented tests and tests reporting abnormal results were associated with higher download rates (p<0.01). Higher volume tests were associated with lower download rates (p=0.03).
In select cases graphical information may be clinically useful, particularly for less frequently ordered tests and in on reports of abnormal results; however, between-test heterogeneity was high. The utilization data presented could be used as a reference point for other laboratories planning on implementing graphical reporting. In many cases graphical elements may add little clinical utility, particularly if these merely reinforce information already contained in text based reports.
Computerized Provider/Physician Order Entry in Anatomic Pathology: A Single Center Experience
Seung Park, MD (firstname.lastname@example.org)1, Anil Parwani, MD, PhD1; Samuel Yousem, MD1; Luke Wiehagen1; Susan Kelly1; Tony Piccoli2; Frank J. Losos, III2; Kara Balatincz, PA (ASCP)1; Liron Pantanowitz, MD1
1Department of Pathology, University of Pittsburgh
Medical Center, Pittsburgh, PA
2Information Services Division, University of
Pittsburgh Medical Center, Pittsburgh, PA
While computerized provider/physician order entry (CPOE) modules in the electronic medical record (EMR) and laboratory information system (LIS) have become common in clinical pathology, their adoption in anatomic pathology (AP) has been slower. The aim of this study is to present our experience with a novel implementation of AP CPOE in a large academic hospital.
APLIS: CoPathPlus (version 3.2, Cerner); EMR: SurgiNet (Cerner Millenium).
Respective modules in the APLIS and EMR were customized in collaboration with the vendor, enabling bidirectional communication of surgical order data to accompany surgical pathology specimens at one hospital (UPMC Shadyside). Surgical nurses entered surgery type and specimen descriptions with orders into the EMR. This order information, along with additional clinical data, was extracted into an APLIS-bound HL7 message. This message was used to accession specimens and populate case information in the APLIS. Adverse event data regarding errors in specimen accessioning were compared for specimens received with and without CPOE. Nurses were then trained in CPOE, and a second round of data collection ensued.
A total of 17924 surgical cases were performed at UPMC Shadyside in a six month interval, 4403 (25%) of which utilized CPOE. There were 92 adverse events reported among CPOE-enabled surgeries (2%) and 127 adverse events among non-CPOE-enabled surgeries (1%). Adverse events due to missing/incorrect patient information, clinical history/procedure and clinician name or other registration issues were completely eliminated in CPOE-enabled surgeries. However, adverse events related to missing/incorrect specimen information, discrepancy between specimen container and paper requisition and truncation of data received due to inappropriate character usage showed sharp increases, together accounting for the vast majority of adverse events reported in CPOE-enabled surgeries. When there was AP CPOE training, adverse events dropped by 75% (Figure).
AP CPOE is feasible and has benefits similar to CPOE in clinical pathology. The increase in adverse events reported in CPOE-enabled surgical cases is largely attributed to the fact that surgical nurses are not adequately trained to properly accession specimens. In conclusion, although AP CPOE is novel at present, further development in this field is anticipated, as the benefits clearly outweigh the challenges.
Impacting Patient Care Through Laboratory Decision Support - Identification and Alert of the Patient's First Troponin-I Increase
Eugenio Zabaleta, PhD (email@example.com), John Burgess, MD, Gregory Eaton, MD, Michael Patterson, DO
MedCentral Health System, Mansfield, OH
MedCentral Physicians requested the laboratory to identify clinically significant changes of Cardiac Troponin results and alert those findings to the caregivers in a timely fashion in order to impact patient care. Cardiac Troponin (cTn) is the recommended biomarker for the detection of myocardial injury or necrosis when assessing acute myocardial infraction by the American College of Cardiology. IT can be used to improve communication for better patient outcomes.
To address the physicians' request, MedCentral has incorporated a Laboratory Decision Support Software called RippleDown® (Pacific Knowledge Systems™). Within this software, rules were created for the purpose of providing clinically relevant alerts based on the outcome of current and previous cTn testing for each patient to warn physicians of potential clinical issues with their inpatients. Consistent application of clinical knowledge through this software enhances communications between the hospital clinical services and has the possibility of decreasing human errors.
Since March, 2010, MedCentral Laboratory is able to identify clinically significant changes of Cardiac Troponin using this artificial intelligence. When the first increase of Troponin-I occurs, RippleDown® will send an alert message to the lab information system. The performing lab professional will call the patient's nurse and communicate the first increase of Troponin-I. The nurse will then phone the physician. The results will also be communicated to the attending and/or consulting physicians through the Hospital Information System under the Alert section through Rules and Workflow software engine.
MedCentral Laboratory compared pre and post implementation data to study the impact of this project. The control group is from December 2009 to February 2010 (pre-implementation); during this period 5195 cTn were performed, 151 would be identified as first cTn increases. The study group was from December 2010 to February 2011; during this period 4755 cTn were performed, 140 were identified as first cTn increases. The most clinically important finding was a significant reduction of length of stay in patients admitted in non-cardiac services with non-cardiac related admitting diagnosis from 7.89 to 6.17 days.
First Increased Troponin-I Alert Project is improving timely communication between hospital's services that leads to better and more effective patient care.