Computer programme to help doctors identify rare diseases

US Veterinarian Craig Carter has written a computer program to help doctors diagnose diseases they rarely see after he contracted a disease that normally strikes cows.

Carter, head of epidemiology and informatics for the Texas Veterinary Medical Diagnostic Laboratory at Texas A&M University, contracted brucellosis through occupational exposure.

Since brucellosis normally affects cattle, none of the doctors Carter consulted recognised the disease; he finally diagnosed himself through analysing his own laboratory blood test results.

‘Many diseases caused by infectious agents or parasites spread from animals to humans present difficult diagnostic challenges because they are so rare in people,’ Carter said. ‘Physicians can overlook, misinterpret or disregard important clues that could lead to an initial diagnosis because they are unfamiliar with these diseases.

‘In my case, I consulted six different doctors, and none of them could tell me what was wrong,’ he observed. ‘That’s when I got the idea that a computer diagnostic database might be helpful in pinpointing the identity of such diseases.’

Carter and fellow Texas A&M scientist Norman Ronald adapted a computer diagnostic program Carter had written for dogs and cats, constructing a searchable database of 223 infectious and parasitic diseases common to animals and humans.

A 1997 limited laboratory trial of the program on several hundred actual cases of Brucellosis in humans was a success, with the results published in the Centres for Disease Control journal Emerging Infectious Diseases, so the authors decided to expand their program. Ronald is currently working on building data for over 300 infectious diseases.

‘The axes of information for each disease include a description of symptoms, clinical and laboratory findings, recommended treatment and a bibliography,’ Carter said. ‘The full Windows-based version will soon be out, available over the Internet or on CD-ROM. Physicians will be able to use the program like a textbook to look up full descriptions of any of the diseases in the database.

‘Or they can input information about a patient, including medical history, findings of their physical exam and results of laboratory tests,’ he continued. ‘The program will then generate a list of possible diagnoses. The great thing is that a doctor can keep inputting additional data and narrow the possible diagnosis.’

Carter is convinced using the computer shortens the time needed to make a correct diagnosis and increases the chances of that diagnosis being accurate.

‘In our 1997 study, on the average it took approximately three minutes of physician interaction with the computer program to construct a differential diagnosis,’ he said. ‘Therefore, this type of computer screening could easily become part of routine history-taking and physical exam of patients, increasing the discriminatory power of clinical history and physical signs and symptoms in suggesting the presence of a rare infectious disease in any particular patient.’

On the web