GIDEON Diagnosis is a computer driven Bayesian matrix for the diagnosis of infectious diseases. It includes virtually all of mankind’s major and minor infectious diseases. In some cases, generic diseases are subdivided for statistical reasons: e.g., herpes simplex infection and herpes simplex encephalitis are considered separately by GIDEON. In other cases, diseases are grouped for lack of sufficient discriminative data to allow for meaningful computer diagnosis. Thus, the designations “Pneumonia – bacterial,” “Meningitis-bacterial,” and “Septicemia” each represent a wide variety of diseases and pathogens.
GIDEON does not include a number of self-defined and obvious infectious diseases, e.g., otitis externa, simple warts, laryngitis, etc. Similarly, GIDEON excludes a number of diseases that do not present as clinical “infection” e.g., subacute sclerosing panencephalitis, Helicobacter gastritis and Kuru.
The major problem in developing an infectious disease diagnosis program is difficulty in obtaining reliable and accurate incidence data. The reporting rate for diseases varies widely by diagnosis as well as country. In many instances, the data used are estimates based on prior years, clinical acumen, and reports from neighboring countries. Care is taken to incorporate the incidence of clinical (as opposed to total) disease acquisition, since our program deals exclusively with the symptomatic patient. Since Bayesian analysis requires that all possibilities be included in its mathematical formulae, the database also includes a number of obvious and self- defined infections, such as bacterial pharyngitis, cellulitis, and wound infection.
A second difficulty in any diagnostic program is the reliability of user input. The accuracy of clinical input is only as good as the accuracy of history taking, physical examination, and laboratory testing. In some instances, clinical observations may be fictitious or unrelated to the present illness.
The accuracy of GIDEON rests on four basic assumptions:
- Reporting statistics for individual diseases are reasonably accurate (or at least consistent within any given country). Numerical data incorporated into this program consist of the published or estimated incidence/prevalence of clinical disease.
- The user’s clinical input is precise and relevant. Questionable or borderline findings should not be used.
- Clinical data are assumed to result from a single infectious disease. A variety of non-infectious diseases may also produce fever, leukocytosis, and other signs of infection. Slow-virus and retroviral infections (other than AIDS) are also excluded, since such disorders are rarely encountered as clinical “infectious disease problems.”
- The patient is a citizen or local resident of the country in question. Incidence data for tourists and expatriates may vary from those of the indigenous population. In some cases, the country of acquisition may not match the country of residence. A careful birth and travel history should be elicited.
GIDEON chooses signs and symptoms on the basis of their discriminative power. Similarly, GIDEON includes generic diseases when group similarity is beyond the scope of this program. For example, GIDEON will distinguish between Mycoplasma, Legionella, Viral, Chlamydial, and other pneumonias, but not between pneumococcal, Haemophilus, and Klebsiella pneumonias.
Often, when dealing with an infection acquired in an “exotic” country, the clinician is primarily interested in accessing a comprehensive differential diagnosis list. In a blinded multicenter study of 495 cases, GIDEON displayed the correct diagnosis in 95%, and ranked this diagnosis first in its differential list in 75% of cases.