Objective. This article considers the possibility that using secondary diagnoses extracted from hospital medical records of patients would modify the diagnosis related groups (DRG) allocated to a case and the Charlson comorbidity index (CCI) calculated from data in the Brazilian hospital information system. Design. This study used two databases: the administrative database of the Brazilian health care system which consists of claim forms abstracted from medical records and is primarily linked to reimbursement, and the medical records that correspond to a sample of claims forms of which the first data source is composed. Changes in DRG were tested by analyzing percent of agreement and the κ index. Logistic regression was employed to evaluate the impact of using CCI scores. Settings and patients. This study is based on a sample of claim forms and medical records (n = 1331) from a number of private acute-care hospitals which had contracts with the municipality of Rio de Janeiro in 1986. Results. Use of information on comorbidity shown in medical records caused changes both in the classification of cases into DRG and in the scores of the CCI. The impact of restrictions on the number of secondary diagnoses in the Brazilian administrative database is comparatively more important for the CCI than for DRG allocation since the Charlson method is based on an additional model where every case of comorbidity is taken into account cumulatively for the final score. Conclusion. These findings indicate the importance of taking a number of measures to improve the quality of information systems in order to increase their potential role in the evaluation of Brazil's health services