An assessment of the Nam Pehchan computer program for the identification of names of south Asian ethnic origin

Background An assessment was made of the usefulness and accuracy of a computer program for the identification of the south Asian population through the classification of names on a disease register. Methods The computer program, Nam Pehchan, was used to classify names as either south Asian or non south Asian. The results were compared with a reference standard, which combined use of the program with visual inspection. The latter was facilitated by a computer-generated dictionary of common non south Asian names. The data set consisted of 356 555 cases of incident cancer (ICD9: 140-208) registered between 1990 and 1992 by Thames, Trent, West Midlands and Yorkshire cancer registries. Results Nam Pehchan classified 5506 cases as south Asian. Visual inspection identified 2024 false positives (36.8 per cent of all cases identified as south Asian by Nam Pehchan) and 363 false negatives (9.5 per cent of those identified by the reference standard). Compared with the reference standard, Nam Pehchan had a sensitivity of 90.5 per cent and a positive predictive value of 63.2 per cent. Conclusion The Nam Pehchan program quickly identified a high proportion of the names classified as south Asian by the reference standard, but the high false positive rate means that the program alone is not an adequate single strategy. The time-consuming process of inspection of program negatives for large data sets can be substantially reduced by comparison with dictionaries of common non south Asian names.