skip to content »

Structural and Computational Biology and Molecular Biophysics

Houston, Texas

A BCM research lab.
Structural and Computational Biology & Molecular Biophysics
not shown on screen

Elmer V. Bernstam, M.D., MSE

Elmer V. Bernstam, M.D., MSE

Professor, Biomedical Informatics

SCBMB Executive Committee Member

University of Texas-Houston


  • M.S., 2001 Stanford University Medical Center (Biomedical Informatics)
  • MSE, 1999 University of Michigan College of Engineering (Computer Science and Engineering)
  • M.D., 1995 University of Michigan Medical School (Integrated Medical-Premedical (INTEFLEX) program)
  • BSE, 1992 University of Michigan College of Engineering (Computer Engineering)
  • B.S., 1992 University of Michigan College of Literature, Science and the Arts (Biomedical Sciences and Psychology)

Research Interests:

Medline QBE (Query by Example)

Medline is a database of medical literature citations maintained by the National Library of Medicine. Currently Medline contains roughly ten million references. Although Medline is indexed using a controlled vocabulary (MeSH: Medical Subject Headings), studies of Medline searching reveal less than perfect results. Published recall and precision rates vary but are generally in the 30 - 70 percent range. Studies have shown that, in general, MeSH based queries return better results than keyword searches. Unfortunately, searching using MeSH requires familiarity with MeSH and experience formulating queries.

Medline QBE is an interface to Medline which allows query refinement by selecting one or more articles from the original search that the user deems particularly relevant. Further, the interface allows easy implementation and testing of novel Medline access strategies. in this way, it is a convenient research workbench for Medline information retrieval.

Dr. Bernstam's current projects focus on using citation analysis to improve Medline information retrieval. Numerous algorithms exist to select articles according to similarity to a query or topic, published in a certain journal or written by a given author. However, few if any algorithms exist that identify the most important information on a given topic. As the sheer volume of the biomedical literature increases, it becomes increasingly difficult to identify the key papers in a specific area. In many ways, this is the same problem faced by web surfers. Therefore, we are applying web algorithms to the biomedical literature to identify the key articles that are relevant to a given query.

Quality and Accuracy of Online Health Information

Despite the potential of the Internet to improve health care delivery, the quality of online health information is problematic. A variety of quality criteria have been suggested in an attempt to help consumers identify misleading, inaccurate or harmful information. Objective quality criteria which offer a limited number of options are particularly promising since they are easier to assess. For example, it is easier to determine whether an author is identified than to determine whether the author is qualified. However, even seemingly objective quality criteria have proven unreliable without specific operational definitions. Further, limited evidence supports the claim that these criteria, known as “technical criteria,” actually filter out problematic health information. The few studies that have attempted to evaluate technical criteria have produced conflicting results. If harmful information can be effectively identified using objective quality criteria, this should be publicized. If, on the other hand, currently available quality criteria cannot identify potentially harmful information, then we should caution consumers and work on finding other ways of identifying problematic information online.

Guideline Compliance Project

In this project we investigate clinical decision support in ambulatory care. Multiple studies have demonstrated variation in clinical practice that cannot be scientifically justified. Therefore, attempts to standardize approaches to specific clinical scenarios (clinical practice guidelines) have been published in the print literature by a variety of organizations. However, these have had little effect on changing clinician behavior. If, however, advice based on these guidelines could be incorporated into the clinical workflow, compliance improves. Therefore, many groups have focused on computerized clinical practice guidelines integrated with physician order entry systems and electronic medical records.

The goal of this project is to explore the meaning of "compliance" with a guideline and to identify how advice can best be delivered by an e-prescribing system in ambulatory care. This project has been funded by the Medical Letter.

Biomedical Informatics Component, Center for Clinical and Translational Sciences at UT-Houston

UT-Houston, in partnership with MD Anderson Cancer Center and Memorial Hermann Hospital System was fortunate to receive one of the 12 first-round NIH Roadmap Clinical and Translational Science Awards (CTSA). The Center for Clinical and Translational Sciences at UT-Houston is funded by our CTSA and includes a number of "core components," one of which is biomedical informatics. I am the co-director of our CTSA biomedical informatics component, represent our site on the national biomedical informatics steering committee and am responsible for daily operations.

The long-term goal of this project is to transform clinical and translational research. Within the biomedical informatics component, we are working toward data sharing within the collaborating institutions and with other CTSA grantee institutions. See the CCTS informatics page for more information and links to publications and presentations.

Selected Publications:

  • Hersh WR, Weiner MG, Embi PJ, Logan JR, Payne PR, Bernstam EV, Lehmann, HP, Hripcsak G, Hartzog TH, Cimino JJ and Saltz JH. Caveats for the Use of Operational Electronic Health Record Data in Comparative Effectiveness Research. Med Care, [Epub ahead of print] (2013). PubMed
  • Joffe E, Byrne MJ, Reeder P, Herskovic JR, Johnson CW, McCoy AB, Sittig DF and Bernstam EV. A benchmark comparison of deterministic and probabalistic methods for defining manual review datasets in duplicate records reconciliation. J Am Med Inform Assoc, [Epub ahead of print] (2013). PubMed
  • McCoy AB, Wright A, Kahn MG, Shapiro JS, Bernstam EV and Sitting DF. Matching identifiers in electronic health records: implications for duplicate records and patient safety. BMJ Qual Saf, 22(3):219-24 (2013). PubMed
  • Herskovic JR, Subramanian D, Cohen T, Bozzo-Silva PA, Bearden CF and Bernstam EV. Graph-based signal integration for high-throughput phenotyping. BMC Bioinformatics, Suppl 13:S2 (2012). PubMed
  • Goodwin JC, Johnson TR, Cohen T, Herskovic JR and Bernstam EV. Predicting biomedical document access as a function of past use. J Am Med Inform Assoc, 19(3):473-8 (2012). PubMed
  • Herskovic JR, Cohen T, Subramanian D, Iyengar MS, Smith JW and Bernstam EV. MEDRank: using graph-based concept ranking to index biomedical texts. Int J Med Inform, 80(6):431-41 (2011). PubMed
  • Bernstam EV, Smith JW and Johnson TR. What is biomedical informatics? J Biomed Inform, 43(1):104-10 (2010). PubMed

For more publications, see listing on PubMed.

Contact Information:

Department: Health Information Sciences and Internal Medicine
Address: 7000 Fannin, Suite 600
Houston, TX 77030
Phone: 713-500-3901
Fax: 713-500-3929

E-mail this page to a friend