Background: Database on hospital records like discharge data, birth and death certificates are widely used for epidemiological and research studies. However there are a very few validation studies on these data. The aim of this study was to validate and assess the accuracy of the ICD 10 database on congenital anomalies in the state of Penang. This study was carried out for three years, from 2002 to 2004. Methods: The list of cases coded under the general coding “Q” was extracted and approximately 30% of cases were randomly selected from the list. Medical records for the selected cases were checked and discrepancies for the diagnoses between the medical records and the ICD 10 data base were recorded for three years. Verification was done for basic demographic variables and the coding of the diseases. Discrepancies, sensitivity and specificity were calculated. Results: The ICD 10 database for congenital anomalies are classified into two types: Type 1 and Type 2. Discrepancies on demographic information were found among the age of patients (babies with congenital anomalies). In Type 1, there was a discrepancy of about 0.02 % to 0.05% probability that a congenital anomaly case can be recorded as non congenital anomaly in the ICD 10. In Type 2 there was a discrepancy that a non-congenital anomaly was classified as congenital anomaly and this ranged from 26.7% to 50.0%. The sensitivity ranged from 96.85% to 97.98%, thus it can be concluded the ICD 10 database is highly sensitive while the specificity ranged from 50.00% to 78.57 %. In other words the ICD 10 is not accurate when classifying the non- congenital anomaly cases. A fair percentage of non-congenital anomaly cases were classified as CA in the ICD 10 database. Conclusion: Even though hospital databases are used as a baseline data for a number of research and epidemiological studies it cannot be used at face value. Validation of these data is necessary before any conclusions can be drawn or intervention measures are undertaken.