O'Brien TF. 1998.
Towards coordinated surveillance of antimicrobial resistance:
the need to merge surveillance databases. APUA
table of contents
|Towards coordinated surveillance of antimicrobial
resistance: the need to merge surveillance databases
Thomas F O'Brien, MD
Brigham & Women's Hospital and Harvard Medical School, Boston Massachusetts, USA
What are we surveying?
Surveillance of resistance to antimicrobial agents ultimately seeks to observe and explain the emergence and dissemination
of hundreds of tiny elements (genes that express resistance to antimicrobial agents) through an enormous universe
(the populations of bacteria on humans and animals throughout the world). It is a huge and intricate problem.
Populations of bacteria pack three billion years of evolution into a bewildering mixture of strains and species,
which are invisible, lack boundaries and smell bad. The resistance genes in bacteria are made variably mobile by
many recombinational tricks of the genetic vectors that carry them. Since a human or animal host may carry more
bacteria than there are people in the world, the spread of a resistance gene through the bacteria of even one host
may be more complex than the pandemic of an influenza virus through the people of the world.
Epidemics of resistance genes are further complicated by being non-random or passive. The emergence of a gene expressing
resistance to an antimicrobial agent, its move from one genetic vector to another, movement of that vector then
from one bacterial strain to another, or movement of a resistant strain from one host to another are all driven
by exposure to that agent. This process is further complicated by the tendency of one resistance gene to become
linked to another gene which is expressing resistance to second agent. As a result of this linkage, either agent
will thereafter drive the spread of both genes (1).
Why are we surveying?
While antibiotic usage and infection control efforts without detailed surveillance have led us to the current understanding
of resistance, we might do better if we knew a lot more about where the resistance genes are globally and locally,
how they respond to different strategies of usage or to different methods of infection control or to both in concert.
Surveillance is thus needed to show us the epidemiology of resistance and how it works amidst all the variables
in the real world in real time where it needs to be managed. Nothing else will do that.
Molecular biology has delineated the resistance genes and their vectors to the nucleotide, but this extrapolates
to practice in an intensive care unit no better than the study of basic physics allows one to determine the year's
weather forecast. Microbial population biology can suggest models for strain selection and observe competition
in vitro, but it can not directly predict what will turn up in your community.
Besides this ultimate goal of understanding the epidemiology of antimicrobial resistance, there are other useful
shorter term objectives for surveillance. For example, the clinician selecting an antimicrobial agent to treat
a patient at a medical center may be helped by a listing of the percentages of isolates of different species which
its laboratory has categorized as susceptible to a panel of antimicrobials. Such lists also help pharmaceutical
companies in marketing their agents. Listings by patient unit may help infection control workers detect problems.
How is surveillance done?
Surveillance systems can be seen to have two major variables: namely sampling and site of testing. Active sampling
obtains defined samples from a defined population of hosts often for susceptibility testing of a defined species,
such as nasopharyngeal cultures of children in day care centers for testing of pneumococci. Passive sampling utilizes
isolates from specimens collected for other reasons, often sent to clinical laboratories by clinicians seeking
to identify and test bacteria infecting their patients. Sites of testing may differ depending on whether the system
accepts test results obtained by the laboratories that collected the samples or if they forwards those samples
or some subset of them to a reference laboratory for testing or retesting.
Active sampling accepts cost to gain defined denominators. Reference laboratory testing accepts cost to gain reproducibility.
Cost eventually translates into reduced sample size, often by orders of magnitude and tradeoffs differ for each
type of resistance. Active sampling may be called for to survey a problem in a community or detect carriers in
a hospital. Reference laboratory testing may be needed for tests with difficult measurements but not for most others.
Huge sample size is needed to discriminate epidemiological detail whether for infection control or for tracing
the spread of distinctive resistance phenotypes.
Can surveillance systems be coordinated?
These considerations have a number of implications. If all surveillance systems were linked together we would have
a total sample size that would still be minuscule compared to the enormity and complexity of the problems they
attempt to survey. However different surveillance systems taken together, even when (or perhaps especially when)
they use different methods, may contain information that is complementary and more than additive. Furthermore,
merging the databases of different surveillance systems would not be difficult now that they are in electronic
files. They undoubtedly have in common most of their data fields and dictionaries (e.g. for host identification,
location, demographics and sometimes prior antimicrobial usage, for culture site, for isolate species and for susceptibility
test measurements). Codes and formats will differ, but interface engines can now facilitate translations between
them. Encryption can make patients and institutions anonymous but retain identity of the systems for the weighting
of sampling and testing differences.
Qualified scholars could apply for selective access to the merged database resource. Print publications gated by
the interests of a few authors can convey only a small fraction of the information in one system's database. A
hundred scholars would find a hundred different interrelationships to explore in a database aggregated from multiple
surveillance systems, and would greatly amplify the yield of information from the individual systems at virtually
no additional cost. The world's bacteria collaborate efficiently to speed the dissemination of resistance. The
world's humans should do the same to slow it.
O'Brien TF. 1997. The global epidemic nature of antimicrobial
resistance and the need to monitor and manage it locally. Clin Infect Dis 24(suppl 1):2-8.