Administrative data show high vaccination coverage rates in Brazil, but there is no assessment of the validity and timeliness of dose administration, or whether the vaccination schedule is complete. This study assessed timely and updated coverage rates in children 12 to 24 months of age. This was a longitudinal population-based study in Araraquara, São Paulo State, a predominantly urban medium-sized municipality, using the Juarez System, an electronic immunization registry (EIR). Coverage rates were assessed in 49,741 children born from 1998 to 2013, a period in which five different vaccination schedules were used. Trends were estimated with the Prais-Winsten linear regression method. Updated coverage of the complete schedule varied from 79.5% to 91.3% at 12 months and from 75.8% to 86.9%, at 24 months. Timely coverage (all doses applied at the recommended ages, with no delays) ranged from 53.3% to 74% at 12 months and from 36.7% to 53.8% at 24 months. There was an upward trend in updated coverage at 24 months. The delays in relation to recommended age increased starting at six months and appeared to relate more to age than to the number of doses in the schedule. The proportion of invalid and late doses was lower than in other studies. Despite the increase in the number of doses in the vaccination schedule, the study showed high updated coverage rates and higher timely coverage than reported in the national and international literature; however, more effort is needed to increase timeliness. EIR proved relevant for assessing and monitoring vaccination coverage with more accurate analyses.

OBJECTIVETo describe adverse events following vaccination (AEFV) of children under two years old and analyze trend of this events from 2000 to 2013, in the city of Araraquara (SP), Brazil.METHODSThis is a descriptive study conducted with data of the passive surveillance system of AEFV that is available in the electronic immunization registry (EIR) of the computerized medical record of the municipal health service (Juarez System). The study variables were: age, gender, vaccine, dose, clinical manifestations and hospitalization. We estimated rates using AEFV as numerator and administered doses of vaccines as denominator. The surveillance sensitivity was estimated by applying the method proposed by the Centers for Disease Control and Prevention. We used Prais-Winsten regression with a significance level of 5.0%.RESULTSThe average annual rate of AEFV was 11.3/10,000 administered doses, however without a trend in the study period (p=0.491). Most cases occurred after the first dose (41.7%) and among children under one year of age (72.6%). Vaccines with pertussis component, yellow fever and measles-mumps-rubella were the most reactogenic. We highlighted the rates of hypotonic-hyporesponsive episodes and convulsion that were 4.1/10,000 and 1.5/10,000 doses of vaccines with pertussis component, respectively, most frequently in the first dose; 60,0% of cases presented symptoms in the first 24 hours after vaccination, however, 18.6% showed after 96 hours. The sensitivity of surveillance was 71.9% and 78.9% for hypotonic-hyporesponsive episodes and convulsion, respectively.CONCLUSIONSThe EIR-based AEFV surveillance system proved to be useful and highly sensitive to describe the safety profile of vaccines in a medium-sized city. It was also shown that the significant increase of the vaccines included in the basic vaccination schedule in childhood in the last decade did not alter the high safety standard of the National Immunization Program.

This article describes the use of connectionist and symbolic learning algorithms in the problem of Bankruptcy Prediction. Data about Brazilian banks represented by 26 or I O indicators of their current financial situation were used. The difference among the number of existent examples in the classes of bankrupt and non-bankrupt banks was livened up through the reduction of learning examples of the class of nonbankrupts and the addition of noise samples in the class of bankrupts.

A Deus, pela oportunidade, aos meus pais, Gervásio e Tarcila, pelo incentivo e apoio, às minhas irmãs, Edna e Edilene, pela compreensão, e especialmente à minha esposa, Silvia, por me acompanhar e dar forças durante os momentos mais difíceis.

The use of a linguistic representation for expressing knowledge acquired by learning systems is an important issue as regards to user understanding. Under this assumption, and to make sure that these systems will be welcome and used, several techniques have been developed by the artificial intelligence community, under both the symbolic and the connectionist approaches. This work discusses and investigates three knowledge extraction techniques based on these approaches. The first two techniques, the C4.5 [12] and CN2 [6] symbolic learning algorithms, extract knowledge directly from the data set. The last technique, the TREPAN algorithm [10] extracts knowledge from a previously trained neural network. The CN2 algorithm induces if … then rules from a given data set. The C4.5 algorithm extracts decision trees, although it can also extract ordered rules, from the data set. Decision trees are also the knowledge representation used by the TREPAN algorithm.