The importance of local data bases in medical expert systems: TICITL.

The database of our medical expert system, TICITL, contains the records of more than 15,000 gastroenterological patients. The data was collected over fifteen years (1977-1992) during which the patients were followed for at least three months to establish a final diagnosis. Using a new set of 230 gastroenterological cases, TICITL's first diagnosis was similar to the final diagnosis in 90% of the patients. When compared to foreign medical expert systems (M.E.S.), there is a considerable difference in diagnostic accuracy favorable to the local system. Another local program is also as accurate as TICITL. Consequently, we attribute these results to the database and strongly recommend employing real local patients whenever possible to implement M.E.S. in a new geographical area.

Differential diagnosis between amoebic liver abscess and acute cholecystitis.

The Research Committee of the World Organization of Gastroenterology has gather information regarding the etiology of acute abdominal pain. Seven diseases cover 96% of the causes of this syndrome in many countries of the world, but some geographical variations have been observed. One example of these variations is amoebic liver abscess, present in 5 to 10% of Mexico City patients. Right upper quadrant pain is often present in amoebic liver abscess and acute cholecystitis. Thus, differential diagnosis of these two entities is difficult. Using discriminant analysis and "stepwise" procedures in 100 cases with cholecystitis and a similar number of patients with amoebic liver abscess, we found six variables (symptoms and signs with a significant chi square to distinguish between these two diseases. The symptoms and signs chosen were hepatomegaly, Murphy's sign, duration of pain greater than or equal to 48 hours, previous history of abdominal pain, dysentery, and facial pallor. These variables proved to be better than laboratory test results. With five of these variables it was possible to obtain an accuracy of 92%. Using six variables, if cases of tie (three variables present and three absent) were excluded, accuracy rose to 96%.

Diagnostico de hipertension porta y/o insuficiencia hepatica en pacientes con hemorragia de tubo digestivo alto utilizando analisis de regression logistica / Diagnosis of portal hipertension and or hepatic insufficiency in patients with uper digestive hemorrhage using logistic regression analysis

Se estudiaron 290 pacientes con HTDA, en 261 de ellos se practico panendoscopia 100 tenian hipertension porta y/o insuficiencia hepatica, 161 presentaron otras causas de HTDA. El analisis de regresion logistica de los datos clinicos y de laboratorio de rutina permitieron escoger tres datos primarios: ascitis, esplenomegalia y leucopenia. Asi como tambien tres datos secundarios (se requiere la presencia de dos de ellos para dar le valor): alcoholismo, ictericia, hepatomegalia dura, como los mejores discriminantes para establecer la presencia o no de hipertension porta y/o insuficiencia hepatica. La exactitud diagnostica con el uso de estos atributos fue de 91%. Cuando se hizo el estudio de 41 nuevos casos (20 cirroticos y 21 que no lo eran) la exactitud diagnostica fue de 92,6%