Wednesday Electronic Poster Session

Presented in the Regency Ballroom B,C,D
 
Wednesday, October 10, 2012

 

3:20 – 4:10 pm
Listed in alphabetical order by First Author 
 
Biorepository Informatics and Management
 
Rocky Ackroyd, MHS (ackror@mmc.org)1,
Tim Marshall2
 
1Maine Medical Center/Spectrum Medical Group, MMC
   Tissue Bank, Department of Pathology, Portland,
   Maine
2NovoPath, Princeton, New Jersey
                                   
Content:
Several efforts at the national level have attempted to develop standards and software for the management of biorepositories, however there are few good commercially available systems that address all needs including a comprehensive database infrastructure, a user interface that streamlines data entry, a simple search component for end users and integration into an existing, sophisticated anatomic pathology laboratory information system (APLIS).
 
Technology:
NovoBioBank is a product of collaboration between Spectrum Diagnostic Services (Portland, Maine) and NovoPath (Princeton, New Jersey) a recognized anatomic pathology software system developer. A total of 26 years of collective experience, both in comprehensive tissue banking and software development has led to an intuitive, robust and information centered system, targeted to fill the aforementioned void in the marketplace. 
 
Design:
The software will function as a stand-alone product for BioBank management and also be available as a module directly incorporated into the NovoPath pathology information system.
 
The development is occurring in three major phases:
 
Phase 1: Comprehensive Database infrastructure development with search capabilities, inventory management, report generation and sample tracking.
 
Status:  Beta software Completed and being tested 
 
Phase 2: Client based order entry to include automation of sample selection, sample retrieval, labeling and shipping functionality and tracking of orders with associated billing  functionality.
 
Status:  Currently in development
 
Phase 3: Plans to incorporate and/or reference clinical data in relationship to samples in the inventory. Linkage of whole slide images to Biobank samples.
 
Status:  Future Development

Results:
Beta software is currently being tested for both ongoing collections and retrospective biobanking programs (including paraffin archive material)   Phase one of the project addresses the database infrastructure needs, search capabilities and basic inventory management. 
 
The software has the capabilities of tracking consent status, disease states, sample formats, availability status, immunohistochemistry and molecular test results. Many quality related data points, including sample weights, ischemic time and pathology verification of samples (presence of normal or diseased tissue and tumor percentage when appropriate) are also integral to the software.
 
Conclusion:
NovoBioBank is the first sophisticated and comprehensive informatics system specifically designed for biobanking that is integrated into an existing anatomic pathology LIS and seeks to fill the void in the existing commercial marketplace. The National Institute of Health and The National Cancer Institute have focused on this need in recent years, the result of which was the development of caBig and caTissue although they have seen limited adoption. NovoBioBank is software that is addressing this need and the development of this tool has been led by both a recognized Pathology Information System vendor and an established tissue bank operation.
 
 
Immunohistochemical Evaluation of Malignant Mesothelioma Tissue Microarray (TMA): National Mesothelioma Virtual Bank
 
Waqas Amin MD1(waa8@pitt.edu), Malini Srinivasan MD, MPH2, Sang Yong Song, MD, PhD2 Nancy B. Whelan MS1, Anil V. Parwani MD PhD2, Michael J. Becich MD PhD1
 
1Department of Biomedical Informatics, University of Pittsburgh and 2Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh, PA
 
Content:
TheNational Mesothelioma Virtual Bank (NMVB) has proven to be a valuable tissue and data resource for the mesothelioma research community. The resource has provided hundreds of specimens including fresh frozen, paraffin and tissue microarrays (TMAs). We perform regular quality assurance on specimens that we shared with the investigators. We performed immunohistochemical evaluation of the University Pittsburgh Medical Center (UPMC) TMA with three key antibodies that are used in making the diagnosis of mesothelioma, and also compared the immunohistochemical assessment between manual scoring and image analysis.

Design and Technology:
The UPMC TMA was assessed for the immune-histochemical expression of three markers used in the diagnosis of mesotheliomas: calretinin (N=39), cytokeratin (CK) 5/6 (N=33), and D2-40 (N=37). Immunohistochemistry was evaluated by semi-quantitative (manual) scoring using light microscope (SS) and by automated image analysis (MS). For the semi-quantitative assessment, one experienced pathologist (SS) scored the immunohistochemical expressions for staining intensity and proportion of positive staining tumor cells was assessed. Staining intensity of tumor cells was classified as negative, weakly positive, moderately positive, and strongly positive. For image analysis, the TMA slide was digitized using Aperio ScanScope XTslide scanner (Aperio Technologies, Vista, CA) at 20X magnification. The tumor areas in each histospot on the TMA were manually annotated using Aperio’s annotation software (ImageScope v11.1.2.760, Aperio Technologies). Calretinin and CK5/6 were assessed using Aperio’s Positive Pixel Count Algorithm v9, and D2-40 was assessed using Aperio’s Membrane algorithm v9.
                       
Results:
·         Calretinin staining was seen in both cytoplasmic and nuclear locations. CK5/6 stain was localized to the cytoplasm. D2-40 stain showed only membranous expression in our cases.
·         Based on the pathologist’s scores, calretinin was positive in 31 of the 39 cases (80%), CK 5/6 in 15 of the 33 cases (46%), and D2-40 in 18 of the 37 cases (49%).
·         The percent positive agreement between manual scores and image analysis was 90% (35/39), 94% (31/33), and 95% (35/37) for calretinin, CK 5/6, and D2-40 respectively. There was substantial agreement between manual and automated scores for Calretinin (kappa=0.614) and almost perfect agreement for CK5/6 (kappa=0.879) and D2-40 (kappa=0.892).
 
Conclusion:
Our study confirms that the immunohistochemical staining pattern of mesothelioma in the NMVB UPMC TMA is similar to other studies. Our findings also show that automated image analysis provides similar results to manual scoring by pathologist, and provides a reproducible, objective, and accurate platform for immunohistochemical assessment of biomarker expression.
 
  
Current Workload and Storage Requirements Associated With Whole Slide Imaging
 
Victor Brodsky, MD1(victor.brodsky@med.cornell.edu), Jessica Pizzarello2
 
1Weill Cornell Medical College, New York, NY
2State University of New York, Oneonta, NY
 
Content:
Currently only the institutions that chose to implement whole slide scanning are fully aware of the required time, effort, and cost. In this analysis we survey the per glass slide workload and the scanner speed to enable organizations interested in establishing a scanning service to define the required number of full-time employee hours.
 
Technology:
Aperio Scanscope AT (2012); Console version 101.0.4.413
 
Design:
167 biorepository and 579 breast consult slides were scanned at “40X” magnification (0.22 microns/pixel) and “20X” magnification (0.46 microns/pixel) respectively. The time spent picking up the slides, loading them into the scanner, obtaining snapshots, adjusting the area to be scanned, modifying focal point locations, performing the scan, verifying image quality, rescanning slides, and typing in associated slide metadata was logged.
 
Results:
At “40X” machine scan time took 22.96 minutes per slide. Rescanning of 2.45% of the slides identified during post-scan quality review resulted in 3.07 additional minutes of machine time per slide in the batch. All associated pre- and post-scan manual steps added up to 1.55 minutes per slide, resulting in the total time per slide of 27.58 minutes. The average file size of the “40X” images was 726 megabytes, with an average scanned area of 11,736,466,666 pixels, or 5.68 cm^2. Consequently, 4.86 minutes and 127.81 megabytes were spent per square centimeter of the scanned glass slide at “40X”. At “20X” the average file size was 498 megabytes and the machine scan time took 3.44 minutes per slide, with 5.5% slides needing to be rescanned thereby adding 0.19 minutes of machine time per slide. The manual steps bring the total scan time at “20X” to 5.18 minutes per slide.
 
Conclusions:
A 7 hour work day of 1 full-time employee followed by 16 hours of machine scan time can result in a predicted maximum of 260 slides scanned at “20X” (likely much less due to human fatigue). At 40X the machine scan time will extend to about 5 days. At this rate a maximum of 13,000 to 67,000 slides can be scanned per year, using up 9 to 32 terabytes (at 40X and 20X respectively).
 
 
Effect of Pen Marking on Glass Slides for Whole Slide Image Scanning
 
Ryan A Collins, MD (collinsra2@upmc.edu), Anil V Parwani, MD, PhD, Walid E. Khalbuss, MD, Jon Duboy, Liron Pantanowitz, MD
 
Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh, PA
 
Content:
Whole slide imaging (WSI) is the digitization of glass slides for archival, education, consultation, and diagnostic purposes. This digitization process often requires an autofocus step to accurately capture (scan) microscopic material on the glass slide, followed by a stitching procedure to correctly merge scanned fields into one confluent digital image. In the practice of anatomic pathology (e.g. cytotechnologists screening slides), marker pens are often used to make “dots” on the coverslip to indicate regions of interest. The aim of this study was to determine if such pen marks (dots) interfere with the aforementioned steps in the imaging process.
 
Technology:
Multiple WSI scanners - Aperio Scanscope XT (Aperio Inc, Vista, CA), Hamamatsu Nanozoomer (Hamamatsu Corp., Bridgewater, NJ), Zeiss Mirax (Carl Zeiss Microimaging, Thornwood, NY). WSI viewing software (Aperio ImageScope 10.2, Vista, CA).
 
Design:
Ten glass slides representing different pathology material were selected including 5 surgical pathology cases (H&E stain) and 5 cytology cases (Pap and Diff-Quick stains). All slides had several (6 to 24) pen marks (dots) on the coverslip surface. Pen colors included blue, black, and green. Dotted slides were scanned on three different WSI scanners (Aperio, Hamamatsu, Zeiss). Slides were then cleaned with ethanol to remove all dots and scanned again using the same machines. All slides were scanned at 20X. Digital WSI images were examined for aberrations (focus, stitching, other artifacts), comparing the exact regions (using ImageScope coordinates) between dotted and cleaned slides.
 
Results:
Pen marks on glass slides resulted in stitching errors, focus problems and other artifacts (pixilation and scanner skipping large portions of the slide) to varying degrees when scanned with different WSI instruments (See Table). 
 
WSI Scanner
Stitching errors
Focus problems
Marked slides
Clean slides
Marked slides
Clean slides
Aperio
10(100%)
0 (0%)
4 (40%)
0 (0%)
Nanozoomer
1 (10%)
0 (0%)
4 (40%)
2 (20%)
Zeiss
9 (90%)
0 (0%)
6 (60%)
0 (0%)
Conclusion:
Digitization of glass slides by all tested WSI scanners was adversely affected by the presence of pen marks (dots) on the glass slide surface. The Nanozoomer had the least issues regarding disturbed image stitching and focusing secondary to dotting of glass slides. These data indicate that pen marks should be removed from glass slides prior to scanning. The results also show that imaging algorithms of WSI scanners display varying capabilities of dealing with foreign markings on the glass slide surface.
 
 
The Impact of “Field-of-Gaze” on Digital Pathology Reviews
 
Catherine M. Conway, BSc. PhD1 (catherine.conway@nih.gov),Neil O’Flaherty, B.Eng2, AviRosenburg, MD1, Alina Nicolae, MD1, Gregory Riedlinger, MD1, Marios Gavrielides, PhD2, Stephen M. Hewitt, MD, PhD1
 
1 Laboratory of Pathology, Center for Cancer
    Research, National Cancer Institute, National
    Institutes of Health, Bethesda, Maryland.
 
2 The Division of Imaging and Mathematics, Office of
    Science and Engineering Laboratories, Center for
    Devices and Radiological Health, US Food and Drug
    Administration, Silver Spring, Maryland
 
Content:
With the advancements in whole-slide imaging technology the need for calibrated, high-resolution monitors has become a necessity in the evaluation of digital slides. There is a general misconception that the monitor and therefore the amount of tissue being displayed must be as large as possible. We have investigated the impact the size of the field-of-view has on a pathologists review. In addition, we have observed that the boundary color which surrounds the digital slide may impact immunohistochemical evaluation.

Technology:
Digital slides were captured at 40x magnification using a Nanozoomer 2.0 HT (Hamamatsu, NJ, USA) instrument. Images were viewed using a web-based software system (SlidePath, Ireland). Slides were reviewed on a consumer grade; 27” LED backlit LCD monitor (Samsung), which was calibrated and color managed to international color consortium standards.

Design:
Twenty-five whole sections were Immuno-histochemical stained for Her-2 protein expression utilizing the Herceptest™ (Dako, CA, USA). Four pathologists reviewed the whole sections on the 27” monitor and scored membrane staining intensity according to the package insert. After a two week wash-out period the pathologists re-scored the 25 cases. However rather than using the entire 27” monitor the area of the screen displaying the slide was greatly reduced, to be comparable to the field-of-view observed under a 20x objective. Images were also displayed with a variation in background color (black, grey and white).

Results:
We have observed that pathologists do not use the entire computer monitor when reviewing digital slides. They in-fact treat the digital slide as they would a glass slide under a microscope, by moving the field-of-interest into the lower centre of the monitor which is in their field-of-gaze. Provisional results demonstrate there is no substantial benefit to reviewing digital slides on a large monitor. In addition we have found that variability in color surrounding the field-of-interest may increase intra-observer variability.

Conclusion
Pathologists do not need large monitors when performing digital reviews. Typically the area under review at any one time should be comparable to the field-of-gaze observed under a microscope at 20x. However, consistency in color surround and the quality of the monitor will impact observer variability. 
 
 
Visual Ergonomics and Complex
Pathology Reports
 
Philip R. Foulis, MD, MPH1 (philip.foulis@va.gov), Carlos A. Muro-Cacho, MD, PhD1, Leah B. Stickland-Marmol, MD1, Matthew Banas, BSIE2, Steven L. Luther, PhD, MA3, Yao Djilan, MPH, VHA4
 
1Department of Pathology and Laboratory Medicine,
    James A Haley Veterans’ Hospital, Tampa, FL
2Department of Systems Redesign, James A Haley
    Veterans’ Hospital, Tampa, FL
3Consortium for Health Informatics Research (CHIR)
    and the VA Center of Excellence: Maximizing
    Rehabilitation Outcomes, Tampa, FL
4Office of Disability and Medical Assessment (DMA),
    St. Petersburg, FL
 
Content:
Human factors engineering involves the integration of computer science, behavioral science, patient safety and visual ergonomics. It can be used to discover the most efficient way to design electronic pathology reports and to facilitate content extraction. The purpose of a Cancer Protocol is to convey diagnostic and prognostic information to health care providers. Implicit in this objective is the clinician’s ability to efficiently and effectively extract critical information. To date, there is limited literature describing the optimal presentation of complex laboratory results.
 
Technology:
Using an audience response system (Turning Technologies, Youngstown, Ohio), we measured answer accuracy and response time to questions regarding information contained in the CAP Fallopian tube Cancer Protocol, purposefully selected because it is a complex report of a rare neoplasm and unfamiliar to most clinicians.
 
Design:
Eight formats containing all required elements were created with variations in capitalization and justification. Four of the formats included a box containing selected diagnostic and prognostic information, which was duplicated in the text component of the protocol. A variety of individuals in the medical field, including students, cancer registrars and oncologists, were randomly assigned one of the formats and queried about protocol content. We measured the speed of accurate responses to a series of questions to determine the best format.
 
Results:
Respondents correctly answered questions more rapidly when using protocols with boxes, independent of other formatting (capitalization, justification). Furthermore, answers containing limited discrete data (e.g. 5, positive, pT3) are better discriminators of report design. No gender differences were observed.
 
Conclusions:
This methodology, using measurable responses to a set of questions, identifies some favorable design elements to optimize efficiency and accuracy of data extraction. Even within a synoptic report, a summary box highlights critical parameters and provides easy access to information necessary for clinical decision making. In the design of future formats, concisely presented data should be favored. This is the first study that investigates speed and accuracy of data extraction from Cancer Protocols. Human factors engineering concepts should be considered when designing complex pathology reports. 
 
Use of Desktop Sharing Software for Dynamic Telecytology Review of Pap Smears: Results of an International Effort Linking Peru and the United States
 
Nicholas C. Jones1 (ncjones@parnters.org), Erika Escalante2, Brenda J. Sweeney SCT(ASCP)1, Barbara Winkler3, William Tench MD4, Rosemary Tambouret, MD1, Nancy Joste, MD5, Patricia Wasserman, MD6, Ronald N. Arpin III, M.S., SCT (ASCP)1, Nasera Hassan CFIAC7, David C. Wilbur, MD1
 
1 Massachusetts General Hospital and Harvard Medical
    School, Boston, Massachusetts,
2 Cervi-Cusco Clinic, Cusco, Peru
3 Mount Kisco Medical Group, Mount Kisco, NY
4 Palomar Medical Center, Escondido, CA
5 University of New Mexico, Albuquerque, NM
6 Columbia University Medical Center, New York, NY
7 Duke University Medical Center, Durham, NC
 
Content:
The Cervi-Cusco clinic in Cusco, Peru, has a large number of international volunteers who collect, process, and interpret Pap smears who come for one or several weeks at a time but only for 4-5 weeks per year. Although the clinic is now served by a local gynecologic cytology screener, there is no pathologist available for the majority of the time. Weekly web conferencing with desktop sharing was used to provide live images to volunteer pathologists and cytotechnologists in the U.S., who interpreted cases identified as potentially abnormal by the cytology screener. Telecytology interpretations allow for more frequent and timely diagnosis and guide treatment of cervical disease in a population that would not otherwise have access to pathologist interpretations. The accuracy of interpretations rendered was measured by colposcopic and histologic follow-up.
 
Technology:
Digital camera attached to microscope, computers with internet access and software to actively view camera images, and desktop sharing “webinar” program (GoToMeeting, Citrix, Santa Barbara, CA, USA).
 
Design:
A mixture of conventional smears and SurePathTM cervical cytology specimens were interpreted by at least one pathologist using the web conference. Slide and field of view selection was based on the review of the CerviCusco cytology screener.
 
Results:
393 cases have been reviewed by this process from 6/10/2011 to 8/17/12. Twenty-five cases have had colposcopy and surgical biopsy follow-up, of which twenty-four cases show correlation, with one case being an under-interpreted as LSIL with a histologic diagnosis of squamous cell carcinoma. In this case, despite the discrepancy, the abnormal result still led to the patients’ colposcopy, biopsy, and correct treatment.
 
Conclusion:
The results of the telecytology review are promising, showing that the real time digital camera images transmitted over web conference are adequate for interpretation. The outcome of the weekly web conference is leading to improved patient care in Peru. In addition this process has the benefit of providing ongoing educational feedback to the cytology screener in Peru. The study is limited by lack of follow up in the population deemed to be negative and the slow rate of obtaining histologic follow up in the population identified as abnormal.
 
 
Evaluation of Laboratory Information System (LIS) Automated Systemized Nomenclature of Medicine (SNOMED) Coding
 
Arivarasan Karunamurthy MD1 (karunamurthya@upmc.edu), Anil V. Parwani, MD, PhD1, Anthony Piccoli, BS2, Liron Pantanowitz, MD1
 
1Department of Pathology, University of Pittsburgh
    Medical Center, Pittsburgh, PA
2Clinical Department Systems – Information Services
    Division, University of Pittsburgh Medical Center.
    Pittsburgh, PA

Content:
Systematized Nomenclature of Medicine (SNOMED) is a widely used coding system with applications expanding beyond pathology. SNOMED coding has evolved into several versions since its release in the 1970’s, along with changes in diagnostic pathology terminology. Accurate coding is a prerequisite in order to reap the benefits of SNOMED coded information. Most anatomic pathology laboratories rely on their laboratory information system (LIS) to accurately generate SNOMED codes. The aim of this study was to evaluate the accuracy of automated SNOMED coding performed by our anatomic pathology LIS.
 
Technology:
Automated coding was performed by the anatomic pathology LIS (CoPath Plus version 3.2, Cerner) autocoder application (MSM) using SNOMED II version. Manual coding was performed using the printed SNOMED II version (CAP, Illinois, published 1979).
 
Design:
We selected 150 surgical pathology reports that were consecutively signed out. For these cases SNOMED codes generated by the LIS were compared with manual coding using the same SNOMED II version. Comparisons were restricted largely to topography (T), morphology (M), procedure (P) and diagnosis (D) axes. Discordant cases were collaboratively evaluated with the LIS vendor.
 
Results:
There were 224 specimen parts with diagnoses for these 150 cases. In 165 parts (74%) automated and manual coding was concordant. For the 59 discordant cases, the reasons why codes did not match were attributed to absent SNOMED codes in 41 (69%) parts, incorrect codes in 14 (24%) parts, and irrelevant extra codes in 4 (7%) parts. No codes existed for relatively newer diagnostic terms such as thyroid papillary microcarcinoma and sessile serrated adenoma. Examples of incorrectly autocoded cases were due to different terminology (e.g. Barrett’s esophagus autocoded as gastric metaplasia and Helicobacter pylori coded as Campylobacter, NOS), words not used by the autocoder (e.g. benign, lateral, squamocolumnar) or incorrect punctuation (e.g. hyphens). Autocoding modified text (i.e. punctuation corrected) improved autocoding results. 
 
Conclusion:
Autocoding of SNOMED codes by the LIS is intended to increase productivity and reduce coding errors. Autocoding can be optimized by adhering to diagnosis text entry rules (e.g. use of correct punctuation and spellchecker) and pathologist validation of autocoded results at sign out. These data show that the autocoder only correctly coded three quarters of our current surgical pathology cases. The majority of improperly coded cases in our study could be ascribed to formatting that hindered the autocoder and/or possibly that an outdated SNOMED version was employed. Therefore, to keep up with newer diagnostic terminologies and classifications, accurate autocoding of anatomic pathology cases by the LIS should employ current versions of SNOMED (e.g. SNOMED CT).
 
 
Automated Mapping of Laboratory Elements to LOINC Using Previously Established CPT Code Mapping
 
Ellen King, MD1 (kinge@uthscsa.edu), Neil Shah, MD2, Adam Harris2, Teresa Bosler2, Richard Scheuermann, PhD2
 
1University of Texas Health Science Center San
    Antonio, San Antonio, TX
2University of Texas Southwestern Dallas, Dallas, TX
 
Content:
The University of Texas Southwestern is composed of several different hospitals, each with its own individual laboratory. Combined, there are greater than 25,000 unique laboratory elements that are both redundant and lack unity. An automated approach was designed to help unify these elements into a single data warehouse by assigning each laboratory element a unique LOINC identifier (Logical Observation Identifiers Names and Codes).

Technology:
An automated LOINC matching program was written using the Ruby programming language.
 
Design:
The algorithm first clusters the laboratory elements according to name and CPT code. A match is created when the name of the laboratory element is at least 85% similar to a LOINC identifier using the Dice Algorithm. Next, the sensitivity of the algorithm is increased by matching data that would have otherwise been eliminated against a generated list of synonymous test names. (Example: “LDL” and “Cholesterol.in LDL”.) Specificity is increased by linking CPT codes to their respective LOINC identifiers using previously established mapping created by UMLS and Mayo Laboratories. Unmatched data is further screened against the complete LOINC lab panel. If still however, no match is found, the “top 2000” LOINC identifiers are then searched for a possible match. If more than one match is found per lab element, then regular expressions for fluid type, units, and time elements are used to filter the data. Finally, a true match will be defined if there are less than 6 LOINC matches per laboratory element.
 
Results:
Of the original 25,377 laboratory elements, this program matched over 15,000 lab tests (62% of total) with near 80% accuracy.  
 
Conclusion:
The LOINC mapping program created at our institution, has proven to be successful at matching LOINC identifiers to our laboratory elements. Ongoing refinement of the program will lead to improved results.
 
 
Evaluation of the International Academy of Cytology (IAC) Virtual Slide Library
 
Liron Pantanowitz MD MIAC1 (pantanowitzl@upmc.edu), Ritu Nayar MD FIAC2, Manon Auger MD MIAC3, Fernando Schmitt MD FIAC4, Patricia Wasserman MD MIAC5, David C. Wilbur MD MIAC6, Walid E. Khalbuss MD PhD FIAC1
 
1Department of Pathology, University of Pittsburgh
    Medical Center, Pittsburgh, PA
2Department of Pathology, Northwestern University,
    Chicago, IL
3McGill University, Montreal, Canada
4University of Porto, Porto, Portugal
5Acupath Laboratories, Plainview, New York
6Harvard University, Massachusetts General Hospital,
    Boston, MA
 
Content:
In recent years web-based learning has been enhanced by offering digitized slides online, since they are more interactive than static images. As a result, virtual atlases using whole slide images (WSI) are now offered by several pathology societies. Since October 2010 the International Academy of Cytology (IAC) began offering its members access to an interactive virtual slide set. The aim of this study was to evaluate the effectiveness of this virtual cytopathology atlas.
 
Technology:
A ScanScope XT WSI scanner (Aperio, Vista, CA) was employed. Images were stored on a 2TB Storage Area Network server. Hyperlinks to these images were incorporated on web pages created with Joomla 2.5.6 (New York, NY) and images made viewable using Aperio's ImageScope viewer. Web analytics were collected using Nova Quiz 2010.
 
Design:
International authors were invited to contribute cases. Their glass slides were scanned and digitized slides shared online as a virtual slide library at http://www.cytology-iac.org/educational-resources/virtual-slide-library. For each case, the WSI and representative static images were offered with interactive questions and answers. In a subset (n=11) of these cases annotation (“dotted regions of interest”) on the WSI were also provided.
 
Results:
A total of 42 cases were posted including Pap tests and non-gynecological cytology. Table 1 compares the number of viewers’ hits on static images and WSI, as well as the proportion of correct diagnoses they entered for these cases. Correct diagnoses for the annotated cases were on average 38%/case for WSI “without dots” compared to 64%/case for WSI “with dots” and 60%/case for static images.
 
Image type
Number of website hits
Correct diagnoses entered
Static
8144
(average 199/case)
Average 56% per case
WSI
(without dots)
1851
(average 45/case)
Average 40% per case
WSI
(with dots)
628
(average 57/case)
Average 64% per case

Conclusions:
Virtual slide libraries are an attractive educational resource pathology societies can offer online. They promote web-based learning and boost web traffic to their website. These data show that viewers preferentially viewed static images. This may be related to user preference, reluctance to use virtual slide technology and/or possibly limited Internet connectivity. Cytology WSI with annotation performs better than those without annotation likely because they require only interpretation skills from the participants, rather than both screening and interpretive skills for WSI without annotation. Digital images are clearly a great boon to pathology education and their role in e-learning is anticipated to grow.
 
 
 
“Unusual” Phlebotomies: Does Collection Above an IV or Saline Lock Result in Variance in Commonly Tested Analytes?
 
Heidi L. Paulin, MD, B.ScH1 (heidi.paulin@dal.ca), Christopher Naugler, MD FRCPC2
 
1Department of Pathology, Dalhousie University,
    Hailfax, Nova Scotia, Canada
2University of Calgary and Calgary Laboratory
    Services, Calgary, Alberta, Canada
 
Content: 
Traditionally, “unusual” phlebotomies (UPs), such as those drawn above a peripheral IV or saline lock, have been relatively contraindicated due to presumed suboptimal performance. This has been challenged by several studies, but most were small or utilized specific patient populations. Using a large quality assurance database in a broad patient population, we aimed to evaluate the difference between common blood analytes drawn via a UP (above an IV or saline lock) as compared to those drawn by conventional phlebotomy utilizing the patient as their own control.
 
Technology: 
Calgary Laboratory Services is the sole provider of laboratory services in the Calgary, AB, Canada metropolitan area, with a catchment of approximately 1.4 million persons and an annual reportable output of approximately 23,000,000 test results. Characteristics about all UPs performed by our staff between December 1, 2009 and September 30, 2011 were recorded in our Millennium laboratory information system (Cerner Corporation, North Kansas City, MO, USA). Statistics were computed using SPSS version 19.0 for Windows (IBM, Armonk, NY, USA) setting α at 0.05.
 
Design:
We queried our quality assurance database for results of six common blood analytes in patients who had undergone a UP and compared them against results from the same patient taken by conventional phlebotomy within seven days. Results were analyzed by paired t-test and univariate logistical analysis looking for statistically and clinically significant differences between the mean test results and UP type.
 
Results:
We studied a total of 7380 paired IV phlebotomies and 4330 paired saline lock phlebotomies. The mean differences between results of UPs versus conventional phlebotomies were statistically significant for most analytes, but none of the differences were large enough to be clinically meaningful. The coefficient of variation for each analysis was small, indicating that the majority of the samples were close to the mean difference (see table). Conditional logistical regression controlling for age and sex showed variable statistical significance for each analysis, but the mean differences were clinically insignificant without an identifiable pattern.
 
Conclusion: 
We demonstrate that selected routine blood analytes show no clinically significant differences when drawn by UP as compared to conventional phlebotomy.
 
 
Analyte and
UP Type
n
(Pairs)
Mean Difference* (95% CI)
Co
efficient of Variation
   P
Sodium, PIV
3084
0.185 mEql/L
(0.080, 0.290)
0.13%
0.001
Sodium, SL
2325
0.054 mEql/L
 (-0.042, 0.151)
0.04%
0.272
Potassium, PIV
3093
0.022 mEql/L
(0.001, 0.043)
0.56%
0.043
Potassium, SL
2330
0.002 mEql/L
 (-0.016, 0.020)
0.05%
0.802
Chloride, PIV
3073
0.432 mEq/L
(0.300, 0.564)
0.41%
<0.001
Chloride, SL
2325
-0.246 mEql/L
 (-0.364, -0.130)
0.24%
<0.001
CO2, PIV
3062
0.387 mmol/L
 (0.277, 0.497)
1.62%
<0.001
CO2, SL
2320
0.580 mmol/L
 (0.479, 0.681)
2.32%
<0.001
Creatinine, PIV
3068
-0.015 mg/dL
(-0.029, -0.001)
1.16%
0.033
Creatinine, SL
2146
0.006 mg/dL
 (-0.007, 0.019)
0.49%
0.380
Hemoglobin, PIV
3921
-0.197 g/dL
(-0.238, -0.156)
1.80%
<0.001
Hemoglobin, SL
2569
-0.033 g/dL
 (-0.071, 0.004)
0.31%
0.085
 
*Mean Difference = mean unusual phlebotomy result – mean conventional phlebotomy result
 
 
Failure Mode and Effects Analysis for the Surgical Pathology Laboratory
 
Luigi K. F. Rao, MD, MS (luigirao@gmail.com),
Thomas M. Gudewicz, MD
 
Department of Pathology, Massachusetts General Hospital, Boston, MA
 
Content:
The surgical pathology section is an essential, intricate division within virtually all laboratories involved in delivering anatomic pathology services. 
It encompasses several different areas from when the sample is obtained in vivo from the patient through the microscopic evaluation by a pathologist. From patient identification and association to accessioning to gross examination to tissue processing, numerous steps are required before patient tissue can be represented on a glass slide and a diagnosis can be rendered. Inherent to this multistep nature are the potential for errors at any level of the entire process. A variety of guiding principles that have been used within business models have been applied to similar stepwise processes including arguably the most well repudiated philosophy in lean manufacturing, best known perhaps in the setting of car maker Toyota’s push for elimination of waste. Failure mode and effects analysis (FMEA) also has been utilized by industry to aid in decision making, goal prioritizing, and risk avoidance. FMEA was initially developed in the 1940s by the United States military and later adopted by the National Aeronautics and Space Administration in the 1960’s. Further advancement for FMEA was done with the civil aviation sector, followed by helicopter and eventually automobile manufacturing. This gradual and diverse adoption suggests industry application of FMEA is not limited but may be modified to fit a particular entity’s operations and workflow improvement programs. 
 
Technology:
QA Assistant (tm) software, Microsoft Excel program.
 
Design:
Twenty-seven functional areas with distinct physical locations and/or separate tasks were initially found encompassing the anatomic pathology laboratory. FMEA was specifically applied to the biopsy service grossing workflow, with risk priority numbers (RPNs) being assigned.
 
Results:
RPNs determined varied widely with high values associated with sentinel event type instances (e.g. 600 for patient misidentification due to wrong container label) to moderate to low values related to delayed turn-around-times (e.g 350 for a case more appropriately grossed in a different section and 60 for additionally sampled tissue given a new accession number).
 
Conclusion:
FMEA’s potential applicability for the surgical pathology laboratory’s multi-layered, time and labor-intensive processes are immense. From effectively minimizing catastrophic failures and reducing probability of significant errors by determining severity level, occurrence frequency, and detection reliability, FMEA can offer those working in surgical pathology a major and effective tool in improving quality metrics. By enhancing patient safety, lowering errors, reducing wastes and associated costs, FMEA should be regarded as a valuable, structured tool within the cache of all who play an integral role in evaluating and promoting excellence within the anatomic pathology laboratory and its many systemic processes.
 
Retrieving Molecular and FISH Data from Pathology Reports: A Comparison of Synoptic and Natural Language Search Tools in a Commercial Anatomic Pathology Laboratory Information System
 
Geoffrey H. Smith, MD (geoffrey.smith@emory.edu), Shelley Caltharp, MD, PhD, Deborah F Saxe, PhD, Karen Mann, MD, PhD, Charles E Hill, MD, PhD, Alexis B Carter, MD
 
Pathology and Laboratory Medicine, Emory University School of Medicine, Atlanta, GA
 
Content:
Molecular and fluorescence in situ hybridization (FISH) laboratories may use anatomic pathology laboratory information systems (LIS) to report results on specimens also being morphologically examined. Some LIS search tools operate on structured information (e.g., discrete synoptic data), whereas others operate on unstructured information (e.g., free text). In this study, we compared the quality and ease of data retrieval between the synoptic search (SS) tool and natural language search (NLS) tool of our LIS.
 
Technology:
Electronic synoptic reports were built in our LIS (Cerner CoPathPlus v3.3, Kansas City, MO) for molecular and FISH tests performed on anatomic pathology specimens.
 
Design:
Results were recorded in both interpretation (free text) and results (discrete synoptic data) fields. SS and NLS queries were implemented for: (a) BCR-ABL1 translocation by polymerase chain reaction (PCR), (b) BCR-ABL1 translocation by FISH, (c) clonal T-cell receptor gamma chain gene rearrangement by PCR and (d) two or more abnormal results by the myeloma FISH panel. The SS result set was validated by comparison with manual record review over a one-month period. Subsequently, the SS and NLS result sets were compared over a 12-month period for (a), (b) and (c) and a five-month period for (d). NLS criteria were iteratively modified to maximize intersecting cases, except for (d).
 
Results:
During validation, SS returned 142 of 143 cases selected by manual review. The missing case did not have a synoptic in error. During comparison of SS and NLS results sets (n=1928), concordant cases (n=1891) were considered to be accurate. Discrepant cases (n=37) were manually reviewed and characterized as shown in Table 1. After multiple iterations of NLS criteria definition, the analytical performance of both search tools was similar for (a), (b), and (c).
 
Conclusion:
Analytical performance of SS and NLS in our LIS was similar. However, multiple iterations of tedious NLS criteria definition were required to achieve this level of performance. The SS tool required one iteration of criteria definition, and the criteria were subjectively far less complex. Therefore, searches (a), (b) and (c) were easier to implement with the SS tool. Query (d), involving a logical grouping operation, was difficult to implement with both tools.
 
Table 1. Reasons for discrepancies between the SS and NLS tools.
 
Reason for discrepancy
Number of cases
NLS criteria not optimized
13
Unexplained NLS behavior
9
NLS unable to filter cases by type
7
Synoptic report and diagnostic interpretation not consistent
7
Missing synoptic report
1
 
 

Copyright 2009 | Site by UMC Web Team