TY - JOUR
T1 - Comparison of validity of mapping between drug indications and ICD-10
T2 - Direct and indirect terminology based approaches
AU - Choi, Y.
AU - Jung, C.
AU - Chae, Y.
AU - Kang, M.
AU - Kim, J.
AU - Joung, K.
AU - Lim, J.
AU - Cho, S.
AU - Sung, S.
AU - Lee, E.
AU - Kim, S.
PY - 2014
Y1 - 2014
N2 - Background: Mapping of drug indications to ICD-10 was undertaken in Korea by a public and a private institution for their own purposes. A different mapping approach was used by each institution, which presented a good opportunity to compare the validity of the two approaches. Objectives: This study was undertaken to compare the validity of a direct mapping approach and an indirect terminology based mapping approach of drug indications against the gold standard drawn from the results of the two mapping processes. Methods: Three hundred and seventy-five cardiovascular reference drugs were selected from all listed cardiovascular drugs for the study. In the direct approach, two experienced nurse coders mapped the free text indications directly to ICD-10. In the indirect terminology based approach, the indications were extracted and coded in the Korean Standard Terminology of Medicine. These terminology coded indications were then manually mapped to ICD-10. The results of the two approaches were compared to the gold standard. A kappa statistic was calculated to see the compatibility of both mapping approaches. Recall, precision and F1 score of each mapping approach were calculated and analyzed using a paired t-test. Results: The mean number of indications for the study drugs was 5.42. The mean number of ICD-10 codes that matched in direct approach was 46.32 and that of indirect terminology based approach was 56.94. The agreement of the mapping results between the two approaches were poor (kappa = 0.19). The indirect terminology based approach showed higher recall (86.78%) than direct approach (p < 0.001). However, there was no difference in precision and F1 score between the two approaches. Conclusions: Considering no differences in the F1 scores, both approaches may be used in practice for mapping drug indications to ICD-10. However, in terms of consistency, time and manpower, better results are expected from the indirect terminology based approach.
AB - Background: Mapping of drug indications to ICD-10 was undertaken in Korea by a public and a private institution for their own purposes. A different mapping approach was used by each institution, which presented a good opportunity to compare the validity of the two approaches. Objectives: This study was undertaken to compare the validity of a direct mapping approach and an indirect terminology based mapping approach of drug indications against the gold standard drawn from the results of the two mapping processes. Methods: Three hundred and seventy-five cardiovascular reference drugs were selected from all listed cardiovascular drugs for the study. In the direct approach, two experienced nurse coders mapped the free text indications directly to ICD-10. In the indirect terminology based approach, the indications were extracted and coded in the Korean Standard Terminology of Medicine. These terminology coded indications were then manually mapped to ICD-10. The results of the two approaches were compared to the gold standard. A kappa statistic was calculated to see the compatibility of both mapping approaches. Recall, precision and F1 score of each mapping approach were calculated and analyzed using a paired t-test. Results: The mean number of indications for the study drugs was 5.42. The mean number of ICD-10 codes that matched in direct approach was 46.32 and that of indirect terminology based approach was 56.94. The agreement of the mapping results between the two approaches were poor (kappa = 0.19). The indirect terminology based approach showed higher recall (86.78%) than direct approach (p < 0.001). However, there was no difference in precision and F1 score between the two approaches. Conclusions: Considering no differences in the F1 scores, both approaches may be used in practice for mapping drug indications to ICD-10. However, in terms of consistency, time and manpower, better results are expected from the indirect terminology based approach.
KW - Drug indication database
KW - International classification of diseases (ICD)
KW - Semantic mapping
UR - http://www.scopus.com/inward/record.url?scp=84901479579&partnerID=8YFLogxK
U2 - 10.3414/ME13-01-0053
DO - 10.3414/ME13-01-0053
M3 - Article
C2 - 24727865
AN - SCOPUS:84901479579
SN - 0026-1270
VL - 53
SP - 195
EP - 201
JO - Methods of Information in Medicine
JF - Methods of Information in Medicine
IS - 3
ER -