Automating Morbidity and Mortality Summaries for Zoos: Trying to Imagine the Forest from Cross-Sections of Individual Trees
American Association of Zoo Veterinarians Conference 1998
J. Andrew Teare, DVM
International Species Information System, Apple Valley, MN, USA


As we approach the new millennium, continued growth of human populations and shrinking wildlife habitats make it seem likely that, at least during the next few decades, zoos will become the last refuges for more rather than fewer species. Long-term survival of small populations of captive wildlife requires intense management that embraces a multitude of scientific disciplines, including genetics, nutrition, ethology and veterinary medicine. The Species Survival Plan (SSP) committee was the response of the American Zoo and Aquarium Association (AZA) to this need to actively manage, as a single population, a species that is often housed in small groups at a number of institutions that are geographically scattered across North America. Under these conditions, seeing the “big picture” is one of the greatest challenges that we face, simply because most of us get to work directly with only a small fraction of the total captive population.

The ISIS Physiological Normals project was one of the first attempts to provide zoo veterinarians with clinically important information derived from multiple institutions. During the first 15 yr of this project, ISIS collected results on more than 27,000 blood samples contributed by 54 different institutions. The reference ranges calculated from these data provided clinicians with the first published set of “normal” hematology and blood chemistry values for a much greater number of wildlife species than was available in the literature. In 1992, the project began to accept records from the MedARKS software package and in the last 6 yr has added almost 50,000 sample records and increased worldwide participation to over 100 institutions. The next publication, scheduled for August 1998, will make more than 2000 pages of reference ranges available to clinicians on a CD-ROM in HTML (World Wide Web page) format.

Using these calculated reference values to automatically mark any test results that are outside the expected range provides the clinician with some diagnostic assistance, but it is still a broad gap between indicating abnormal test results and arriving at a diagnosis. Closing this gap will require much more knowledge about the common diseases and problems that occur in a particular species. Information about the diseases of captive wildlife is mainly recorded in the literature as case reports, in disease reviews, and in some specialized texts. The medical literature has been systematically reviewed and summarized for only a limited number of species and I am aware of only three species where existing medical records from multiple institutions were reviewed in an attempt to more fully categorize the diseases and problems of a species. Both literature surveys and reviews of medical records are labor intensive and time consuming. Thus, it is a fairly rare event for a zoo clinician to have information on the incidence and prevalence of various problems in a particular species and it is difficult to see how this might change if we continue to rely on manual analysis techniques.

Over the past decade, the MedARKS software program has been adopted as the “de facto” standard for computerized medical records in zoos. Currently there are more than 1.5 million medical records in the MedARKS format, making this the single largest computerized database of medical information on captive wildlife. Unfortunately, it is also a distributed database with those records scattered across 150 institutions in at least a dozen countries and in several languages. If this information could be assembled in one location, it is estimated that it would occupy about 3 gigabytes of disk space. However, it would only be worth undertaking that effort if the information could then be more easily used to advance the field of captive wildlife medicine.

This adoption of a standard format for medical records has allowed experimentation with merging information from a number of institutions into “library” disks that can be accessed through the MedARKS software. Early efforts have included a disk of giraffe immobilization drugs and doses and a disk of carfentanil immobilization records for a variety of species. Attempts to assist SSP medical advisors have also led to limited experiments that consolidate all MedARKS records for an individual species. As an example, the assembled MedARKS records for Malayan tapir (Tapirus indicus) contain 23 animals with problem-oriented medical records (defined medical problems in the format of a “master” problem list). A few minutes work with this disk allows you to tally those problems that are recorded for more than one of these animals (Table 1). While this partially automated analysis of medical records is far easier than the equivalent analysis using paper records, it completely ignores the 54 other tapirs where the MedARKS records were not problem-oriented and for which only text records were available. The question of whether those animals with problem-oriented records accurately reflect the diseases/problems seen in the overall population cannot be answered without a much greater in-depth (and time-consuming) investigation into the medical records.

Table 1. Medical problems in Malayan tapirs as derived from MedARKS records

Medical problem

Percentage with this problem in medical history*

Lameness/foot problems


Lacerations/abrasions/bites wounds


Anorexia/partial anorexia




Conjunctivitis/ocular discharge




Self-inflicted trauma


*Problem occurs at least once in the master problem list for an individual.

As we continue to move from paper record systems to computerized medical records, the potential certainly exists for automatically extracting the common causes of morbidity and mortality for a species. There is even the potential to begin to address issues of actual disease incidence and prevalence within a population. The traditional problem-oriented record with a master problem list is well suited for automated analysis, but, as most clinicians have discovered, it is also more difficult and time-consuming to maintain than simple text records. Automated searching of text records for a series of words or phrases is a relatively simple process and can quickly be used to produce a list of records for further review. However, techniques for automatically extracting critical information from text records and summarizing the causes of morbidity and mortality are in their infancy. Developing effective algorithms to extract morbidity and mortality information from text medical records poses the next great challenge.

In summary, as we enter the 21st century, we will need improved access to the experience and knowledge that reside in our medical record files. Theoretically, computers offer a means to more efficiently tap into this knowledge and the widespread acceptance of MedARKS software by the zoo community provides a common format for sharing our medical experience. The problem-oriented record, with well-defined problems and onset and resolution dates, offers the easiest means to automate the analysis of medical records. Computer assisted extraction of useful information from text records is a much more difficult process. However, the current reliance of many zoo clinicians on text records means that this is a challenge that we will need to face even as we promote the use of the problem-oriented approach to keeping medical records.


Speaker Information
(click the speaker's name to view other papers and abstracts submitted by this speaker)

J. Andrew Teare, DVM
International Species Information System
Apple Valley, MN, USA

MAIN : All : Automating Morbidity and Mortality Summaries for Zoos
Powered By VIN