Free 15 Day Trial Latest Updates Video Tutorials

IDCP 1998

GIDEON: A Computer Program for Diagnosis, Simulation and Informatics in Geographic Medicine

Stephen A. Berger, M.D.

Reproduced from: Infec Dis Clin Prac 1998; 7:383-386

From the Department of Geographic Medicine, Tel Aviv Medical
Center and the Sackler School of Medicine, University of Tel
Aviv, Tel Aviv, Israel

Correspondence to:  Stephen A. Berger, M.D., Geographic
Medicine, Tel Aviv Medical Center, 6 Weitzman Street, Tel Aviv
64239, Israel.  Tel: +972 3 697 3263; Fax: +972 3 613 2892;

Over 230 generic infectious diseases are currently
distributed in a seemingly haphazard fashion among 200
countries.  There are new diseases, reemerging diseases,
disappearing diseases, outbreaks and epidemics.  There are the
disease patterns of India and the disease patterns of Peru,
both of which change continually from year to year. The
infectious diseases of southern India in summer are different
from those of northern India in spring.

Behind these diseases lurk over 1,000 pathogens, each
having its own ecology, disease association and phenotype.  At
least 240 generic drugs and vaccines are currently available
to combat these pathogens, distributed under 2,500 proprietary
names.  Pharmacology, dosage, drug interactions,
contraindications and toxicities vary widely from agent to
agent. The literature of Infectious Diseases is diffuse and
largely outdated.  Textbooks and journals cannot follow the
evolution of diseases, pathogens and drugs in ‘real time.’

In 1988, we initiated a project to record and follow the
literature of Infectious Diseases using computer systems.  A
series of computer spread sheets, data bases and macros were
designed to access, collate and enter data from all relevant
printed and electronic sources dealing with the diseases,
their epidemiology and treatment.  Subsequent developments in
computer technology permitted addition of statistical matrices
for simulation and identification of diseases and pathogens.

A commercial program was released in 1991.  GIDEON
(Global Infectious Diseases and EpidemiOlogy Network) is
designed to diagnose or simulate any infectious disease
worldwide, or to identify any species of bacterium or yeast.
Additional modules follow the global and country-specific
status of all infectious disease; and the pharmacology and
usage of all anti-infective drugs and vaccines.


The Diagnosis module generates a ranked differential
diagnosis based on country of disease acquisition, signs,
symptoms, patient background (eg, age, underlying diseases),
laboratory data, exposure history and incubation period.  The
user may select any number of disease features using mouse or
key strokes.  The resulting diagnosis list is generated by a
Bayesian matrix which examines both disease incidence and
symptom prevalence within each specific disease.

Additional options permit the user to interact with this
list.  For example, one might ask why any additional disease
was discounted, or why a low ranking was assigned to one of
the listed conditions.  A ‘what if’ function is also included
for entry of additional clinical data, or simulation of a
given set of symptoms through several alternative countries.
The latter function is useful in examining the differential
diagnosis for a patient who had traveled through several
countries in sequence.   Printouts of the differential
diagnosis of biliary disease and eosinophilia are presented
for Thailand (figure 1a) and France (figure 1b).  Accompanying
notes on the status of the principal diagnoses for these
countries are as follows:

“Opisthorchiasis in Thailand:   The disease is commonly
reported from the northern and eastern regions. Opisthorchis
viverrini predominates.  Highest metacercaria burdens among
cyprinoid fish (Hampala dispar, Puntius leiacanthus,
Cycloceillchthys armatus) occur during July to January.
Coinfection by Echinostoma spp. is common. Seven million were
estimated to be infected during 1980 to 1981. Infestation is
associated with high rates of cholangiocarcinoma in endemic
areas (36 to 84/100,000/year in Khon Kaen). “

“Fascioliasis in France:  The disease is most common in
Lyon, Bretagne Nord/Pas de Calais and Sud-Ouest. Two cases
were reported from Corsica during 1969 to 1989. Recent rates
of approximately 0.03/100,000 per year are reported. Bovine
infection in Amiens decreased from 12% to 2.5% during 1983 to
1995.  The first modern epidemic occurred in 1956 (over 500
cases). 3,297 cases were officially reported during 1950 to
1983 (but 4,813 cases were registered by 23 hospital
laboratories during 1970 to 1982). 58 cases were reported
during 1982 to 1986 – watercress implicated in 89%, dandelion
and other greens in the remainder. The local reservoir is
Lymnaea (Fossaria) truncatula.”

The program currently contains over 10,000 such country-
disease summaries, which can be printed in hard copy, or
transferred to a word processor for preparation of reports,
lecture notes, etc.


The epidemiology module may be used to access data on the
clinical and epidemiological profile for any of over 230
diseases, ranging from conditions such as the common cold and
pediculosis, to Brainerd diarrhea, metorchiasis and the New
World phleboviruses.  Specific modules are assigned to high-
impact global diseases (AIDS, cholera, malaria, tuberculosis,
yellow fever) and the W.H.O. Training in Disease Research
diseases (Leprosy, Filaria, etc).

The user may access a list of countries endemic for any
specific disease, or explore all diseases for an given country
or the world as a whole (eg, the global status of measles or
onchocerciasis). An additional option generates disease lists
based on epidemiological profile; for example, a list of all
mosquito-borne diseases, vs. the mosquito-borne viral diseases
of Brazil.


The Therapy module profiles the pharmacology and usage of
all vaccines and anti-infective agents (eg, antiparasitics,
anti-retrovirals, antibiotics, antifungals, etc).  Drugs may
be accessed by either generic or proprietary name (2,500 in
the current version). Data include the dosage for adult and
child, toxicity, CSF penetration, dialysis adjustments, drug
interactions, laboratory testing standards and spectrum for
each agent.  Options allow for listing by spectrum for any
pathogen or group of pathogens (eg, what two drugs might be
active for a patient infested by Capillaria, Gnathostoma and
Trichostrongylus); or a list of drugs associated with a given
form of toxicity (eg, drugs which exacerbate porphyria) or
interaction (eg, the three anti-infective drugs which interact
with Viagra c).

The vaccine module presents similar data, as well as
dosing and booster schedules for all vaccines and globulin
preparations, ranging from tetanus toxoid to Argentine
hemorrhagic fever vaccine.


The Microbiology module is designed to identify, simulate
and characterize all bacteria, mycobacteria and yeasts.  The
user may enter any combination of phenotypic reactions, in any
order.  The resulting list of microbial taxa is similar to the
diagnosis list discussed above (figure 2a).  The user may
‘ask’ why additional taxa are not listed, or why a given taxon
is assigned low probability or ‘rare biotype’.  As in other
modules, a hard-copy printout can be generated, or the
organism profile can be saved in a file for addition of
subsequent test results or use in lectures, reports, etc.

Unlike all standard identification systems used in
clinical laboratories (API, Vitek, Microscan, etc) a Bayesian
matrix is used to assign identification of probabilities 1.  In
other words, bacteria are ranked according to best fit to the
specified biochemical profile, as well as the relative
incidence (prior probability) of each species in clinical
material.  For example, the following biotype is typical of
most isolates of both Escherichia coli and Edwardsiella tarda:
Gram negative, motile, facultative bacillus which grows on
MacConkey agar and is positive for catalase, indole, methyl
red, nitrate, lysine decarboxylase and fermentation of glucose
(with gas), maltose and mannose; and negative for citrate,
DNA’se, Voges Proskauer, oxidase, malonate, gelatin
hydrolysis, lipase, urease, phenylalanine deaminase, growth in
KCN and fermentation of adonitol, arabitol, cellobiose and
inositol.  When an organism with these features is isolated
from clinical material, we know intuitively that the
probability of Escherichia coli is far greater than that of
Edwardsiella tarda; however, all standard identification
systems will assign a 50% probability to both taxa – simply
based on biochemical fit.

Since the GIDEON software weighs the incidence of each
taxon in clinical material, a more accurate picture of
probability ranking is generated for these two taxa (figure
2a).  An additional function in GIDEON allows the user to
review additional tests which might resolve the final
identification of such an organism (figure 2b).  Note that
this list is optimized; ie, the most discriminative test (in
this case, hydrogen sulfide production) is ranked first on the
list.  This function also allows the user to prepare custom-
made comparison charts of any 2 to 21 taxa for use in
lectures, reports, etc.  The prior and current nomenclature of
any given taxon is also available, as are a number of human
pathogens which are not yet included in standard texts or
commercial identification systems: Dolisogranulum prigum,
Bisgaard’s taxon 16, Corynebacterium falsenii, Gemella
bergeriae, Prevotella pallens, Cryptococcus macerans,
Mycobacterium heidelbergense, Erwinia persicinus, Abiotropha
elegans, et al.

The reference sources for GIDEON include all printed or
electronic health ministry reports, publications of the World
Health Organization and Centers for Disease Control; as well
as standard texts and peer-review journals in Infectious
Diseases, Clinical Microbiology, Tropical Medicine, Virology,
Antimicrobial agents, etc.  In addition, a Medline review is
performed monthly utilizing a computer macro incorporating all
key terms in the program database.

GIDEON is currently installed at over 1,500 sites in 40
countries: health ministries, W.H.O., military facilities,
Infectious Diseases departments, Microbiology laboratories,
libraries, etc.  The program is sold by subscription, with
quarterly updates.


Two blinded studies have examined the performance of the
GIDEON diagnosis module.  The first found that the correct
diagnosis appeared in the differential diagnosis list for
94.7% of 495 cases (sensitivity) and was ranked first in 75%
(specificity) 2.  The second study of 86 hospitalized patients
demonstrated somewhat lower sensitivity (69%) and specificity
(60%) 3.  Although a self-correcting ‘Why not’ option in this
module can be used to access exclusion criteria for diagnoses
which fail to appear in the differential list, the cases
examined in the second study have since been discarded, and
are unavailable for further analysis.


1. Berger S. Lack of precision in commercial identification
systems.  Correction using Bayesian analysis.  J Appl
Bacteriol 1990; 68:285-289

2. Berger S, Blackman U. Computer program for diagnosing and
teaching geographic medicine. J Trav Med 1995; 2:199-203

3. Ross JJ, Shapiro DS. Evaluation of the computer program
GIDEON for the diagnosis of fever in patients admitted to a
medical service.  Clin Infect Dis 1998; 26:766-767