Using NAPIIA to Improve the Accuracy of Asian

Using NAPIIA to Improve the Accuracy of Asian Race Code in Registry Data Mei-Chin Hsieh, MSPH, CTR Lisa A. Pareti, BS, RHIT, CTR Vivien W. Chen, Ph. D NAACCR Conference, Denver, June 2008

Background n n Overall, Asians have lower risk of developing cancer than non-Hispanic whites and blacks For certain types of cancer, such as liver and stomach, Asians have higher incidence rates than other races For a registry with small numbers of Asians, even a few miscoding on these minority races could potentially bias the estimation of incidence rates To ensure Asian races are coded correctly, the Louisiana Tumor Registry implements NAPIIA into its routine data quality procedure

Purpose n n To demonstrate how the NAPIIA can improve the coding accuracy on Asian races To find the misclassification on Asian groups

Asian Population in Louisiana From Census 2000: 54, 022 (1. 208%) Race Count Percent 04: Chinese 7, 474 13. 8 05: Japanese 1, 519 2. 8 06: Filipino 4, 504 8. 3 08: Korean 2, 876 5. 3 09: Asian Indian/Pakistani 9, 050 16. 8 24, 358 45. 1 11: Laotian 1, 334 2. 5 12: Hmong 14 0. 0 13: Kampuchean 310 0. 6 14: Thai 680 1. 3 1, 903 3. 5 10: Vietnamese 96: Asian, NOS

Methods and Approach n n n Data source: Louisiana Tumor Registry Cases diagnosed in year 1995 to 2005 with race 1 (NAACCR item 160) coded to any Asians, other race, unknown race, or non. Asian race with birthplace in Asian country were selected Converted race 1 to 96 (Asian, NOS) and applied NAPIIA on records

Methods and Approach n n New Asian codes assigned by NAPIIA were compared with original race 1 and manually reviewed when the assigned Asian codes were different from the original race 1 codes Research sources utilized for the review: Abstract Text, Accurint Voter Registration, Online Death Certificate, Online Medical Records, and contact hospitals as last resort

Results n Out of 221, 732 cases diagnosed between years 1995 and 2005, 1, 711 (0. 77%) eligible cases were run through the NAPIIA n n Non-Asian race with birthplace in Asian country: 58 (3. 4%), white 55 and black 3 Specific Asian codes: 837 (48. 9%) Asian NOS: 238 (13. 9%) Unknown race: 578 (33. 8%)

Results n n 21. 8% (374) of cases were identified with race coding differing between original race 1 and NAPIIA Comparisons n n n Original race vs. NAPIIA vs. reviewed race Original race vs. reviewed race

Results: Comparing Original Race with NAPIIA § 767 (44. 8%) cases had original race unchanged § 570 (33. 3%) cases had same Asian race codes § 374 (21. 9%) cases (highlighted in yellow and blue) had inconsistent race code between original race and NAPIIA, which required manually reviewed

Results: Comparing Original Race with NAPIIA

Results: Distribution of 374 Cases with Inconsistent Race Codes

Results: Comparing Reviewed Race with NAPIIA § After manually reviewing, of the 374 inconsistent race code 254 (67. 9%) were identified with same race code as assigned by NAPIIA § Of the 89 Filipino codes assigned by NAPIIA, 48 (53. 9%) cases were white (mainly Hispanic) after review

Results: Comparing Reviewed Race with NAPIIA

Results: Comparing Original Race with Reviewed Race § Out of 374, only 34 (9. 1%) cases were initially coded correctly § 46. 3% (19 out of 41) of Asian Indian/Pakistani were recoded to Vietnamese after review

Results: Final Race Categories After Review n n n White: out of 52, 37 were actually Asian Black: all remained as black (due to incorrect birthplace) Asian races: out of 91, 79 were misclassified or miscoded Asian NOS: out of 167, 163 were able to be classified with a more specific Asian race Unknown race: § out of 61, 60 were more specifically classified to a correct race code § 42 (70%) out of 60 were white

Conclusions n n Through this exercise, we were able to reassign the correct race on 340 (90. 9%) cases out of 374 cases reviewed Miscoding was one of the main reasons for misclassification of race 1, other reasons included multiple races and code transposition § Miscoding: code 10 to 09, 04 to 05 § Patient with multiple races: white and Asian Indian § Code transposition: code 10 to 01

Conclusions n n n NAPIIA was able to more accurately identify Vietnamese race group compare with other Asian race groups Filipino race code had the least improved accuracy among race groups after NAPIIA Reduce the percentage of unknown race and Asian NOS n n Unknown race: 0. 26% to 0. 23% Asian NOS: 0. 11% to 0. 07%

Conclusions n n n For a registry with small proportion of Asian cases, NAPIIA seems to be an excellent tool to improve the race coding accuracy on Asian groups NAPIIA also can be applied to race codes other than Asian NOS to enhance registry data quality (with review) A potential additional benefit of using NAPIIA for data quality control is the identification of cases with incorrect birthplace

Recommendations n n n Double check race codes to make sure you coded what you intended If race is known, document the race information in the PE text field. For example, Filipino male If race information is obtained from death certificate or other sources, make sure the corresponding NAACCR race code is coded

Recommendations n Factors that could improve the NAPIIA’s performance § Correct Spelling on last and first name § Maiden name § Birthplace