[Evaluation of data quality from the information system on live births in 1997-1998]

To verify the concordance between data routinely collected by the Sinasc (Information System on Live Births) and data obtained by a cross-sectional survey (gold standard).

The survey was performed in a sample of 2,831 hospital deliveries in ten maternity hospitals of the municipality of São Luís, Brazil. The sample frame represented about 98% of hospital births. Sinasc's data was compared with the survey data using a computerized linkage program. Sinasc's hospital births coverage, percentage of missing or unrecorded data and consistency between Sinasc 's and the survey's data were analyzed. Concordance was measured using the kappa indicator for qualitative variables and the intraclass correlation coefficient for quantitative variables.

The estimated Sinasc coverage was 75.8% (95% CI: 73.3%-78.2%). Five Sinasc's data fields showed good reliability: birth weight, newborn sex, hospital of birth, type of delivery and maternal age. Low birth weight rates estimated by the two data sources were similar and the concordance was high (kappa=0.94). However, preterm birth rates were higher in the survey (11.2%) than according to Sinasc 's data (1.7%) and the concordance was low (kappa=0.09).

Estimated Sinasc coverage was low. Low birth weight rates from Sinasc's data seem to be valid and reliable. Preterm birth rates from Sinasc's data are underestimated.