754 User Centred Networked Health Care A. Moen et al. (Eds.) IOS Press, 2011 © 2011 European Federation for Medical Informatics. All rights reserved. doi:10.3233/978-1-60750-806-9-754 The Revision of the Korean Classifications of Health Interventions Based on the Proposed ICHI Semantic Model and Lessons Learned

Boyoung JUNGa, Chaeyoung JUNGb, Jean Marie RODRIGUESc,d,e,1, Cédric BOUSQUETc,d , Anand KUMARc, Pierre LEWALLEc,, Béatrice TROMBERT PAVIOTc,d , Hoonshik YANGf, Sukil KIMa a The Catholic University of Korea ,Seoul Korea b University of Utah, Salt lake City USA c University of Saint Etienne, CHU, Department of public health and medical informatics ,Saint Etienne France dINSERM UMR 872 Eq 20 Paris, France eWHO Collaborating Center for International Classifications in French Language, Paris, France f College of Medicine Chung-Ang University, Seoul Korea

Abstract. The Korean Medical Association and the Health Information Review Agency have decided to re-engineer the different Korean coding systems of health interventions based on a proposed ontology framework defined in 2010 for the prospective International Classification of Health Interventions (ICHI). The authors present the interim report of the project focused on this model: 5,338 procedures of the Korean version of ICD9-CM 5,150 procedures covered by Korean health insurance and 6,619 uncovered procedure labels were processed with the participation of 8 coders and 310 medical doctors. As of 28th January 61.8% of data was processed. The ontology framework model itself was not enough to represent all the labels when the preliminary data from obstetrics and gynecology was explored. However, when modified with 7 notations, it was possible to assign each label of ICD 9 CM Volume 3 and 30 % to 57 % of specific Korean interventions to the semantic model.

Keywords. Health Interventions, Classifications, Ontology, Semantic Model.

1. Introduction

The WHO Network for the Family of International Classifications agreed in 2006 that a prospective structure for an International Classification of Health Interventions (ICHI) should be explored based on the CEN/ISO ontology framework named Categorial Structure [1]. A backbone was approved in 2008 as a support to harmonize the existing national classifications and to provide a basic system for the countries which have not

1 Corresponding Author: JM Rodrigues, CHU de St Etienne, SSPIM, Chemin de la Marandière, 42 270 Saint Priest en Jarez, France, [email protected] B. Jung et al. / The Revision of the Korean Classifications 755 developed their own classification systems of interventions [2] [3]. Since then, the work is continuing towards implementation [4]. The semantic model has three semantic categories or axes: action, target and means. The action is the main axis defining the key of the procedure. The target includes anatomic structures, body parts, functions, or individual where the action is applied. The means refines the action by showing how the action is applied to the target. For each semantic category, preliminary values sets have been defined [4] following the mapping exercise between existing coding systems of health interventions and the semantic model [5]. Their definitions are presented in another conference full paper. The Korean health insurance is the national health insurance of the Republic of Korea with a fee-for-service payment system. The interventions listed as covered procedures of health insurance are reimbursed to the payer. The other interventions named non covered are paid by the out-of-pocket money of the patients. The covered procedures and the non covered procedures have different hierarchies and coding schemes. The legacy coding system for covered procedures is fairly satisfactory. There are, however, a few caveats. If a non covered procedure became a covered procedure then the procedure is deleted from the non covered coding system and entered into a revised coding system of covered procedures. The Korean Classification of Procedures was first built in 1994 for health insurance claims. To overcome the caveats, the Korean Medical Association (KMA) and the Health Information Review Agency (HIRA) decided to revise it based on the proposed ICHI semantic model till the end of 2011.

2. Material

The data comes from 3 sources. 1. 5,338 procedures labels of the Korean version of ICD9-CM volume 3 were used as the backbone of the classification. 2. The legacy classification of health interventions for health insurance in Korea has 5,150 procedures labels (covered procedures). It is the main target data to be included in the new classification. 3. For the non covered procedures, 44 university hospitals were requested to submit the data. 18 hospitals submitted the data and after the inspection only 6,619 procedures labels are processed. The data from other hospitals will be processed after they meet the inspection criteria. The total number of labels is 17649.

3. Method

The whole process is composed of data collection, data cleaning and assignment of procedures to the 3 semantic categories, validation and hierarchical rearrangement according to the model. The current stage is in the assignment of procedures to the 3 semantic categories . The validation and rearrangement will be done before the end of year 2011. A bilingual web site (Figure 1) with Microsoft SQL server has been built to process the data. 8 graduate students trained in medical informatics have assigned each procedure label to the 3 semantic categories. 5 of them are nurses, 2 are health 756 B. Jung et al. / The Revision of the Korean Classifications information managers and one is a dental hygienist. Twenty four academic societies are requested to validate the assignment to the semantic categories on the web. 310 doctors registered to participate in the process.

Figure 1. Collaborative web tools for review

Figure 2. Hierarchical rearrangement of items according to 3 semantic categories axes

535 procedures labels including 306 ICD9-CM procedures labels related to obstetrics and gynecology were preliminarily analyzed and went through the hierarchical rearrangement (Figure 2). B. Jung et al. / The Revision of the Korean Classifications 757

4. Result

On 28th Jan. 2011, 10,812 items out of the total 17,649 have been assigned to the 3 semantic categories axes (61.3%) (Table 1). The ICD9-CM volume 3 labels were finalized earlier than others to see how usable it was.

Table 1. The progress in the assignment of 3 axes according to sources (as of 28th Jan). Sources No. of Items with Completion(%) Assignment/No. of Items Korean ICD-9-CM 5,338/5,338 100.0 Covered (Legacy) items 1,722/5,692 30.3 Uncovered items* 3,752/6,619 57.0 Total 10,812/17,649 61.3 * collected data only Some limitations of the proposed ICHI semantic model were assessed during the work. Several notations were introduced to overcome them: (Table 2). Table 2. Examples showing notations that were used to modify the 3 axes semantic model.

Case Item Axes No. Code Description Target Means Action 1 73.6 : Open Incision Perineum 2 65.62 Other removal of remaining Open Removal and tube & Ovary 3 66.93 Implantation or replacement Device: Open Implantation of of prosthesis of fallopian tube Prosthesis>Fall device|Change opian tube 4 66.62 Salpingectomy with removal Fallopian Open Excision&+Re of tubal pregnancy tube&+Fetal or moval embryonic structure 5 73.21 Internal and combined Fetal or Per Reposition: version without extraction embryonic Orifice/Tr Internal structure ansorifice version and combined version &- Extraction: Delivery 6 75.92 Evacuation of other Vulva|] Open Drainage: hematoma of vulva or vagina Hematoma Evacuation 7 66.93 Implantation or replacement Device: Open Implantationof of prosthesis of fallopian tube Prosthesis>Fall device|Change opian tube If the granularity of the one semantic category axis is coarser than the granularity needed by the label, : is appended at the end of the semantic category value set and after the symbol the more fine granularity value is registered (target in case 1). When more than a target or an action were needed, & was inserted denoting and (target in case 2). | was used denoting or where more than two options were available in an axe (action in case 3, target in case 6 and action in case 7). 758 B. Jung et al. / The Revision of the Korean Classifications

When one item was associated with the other item then &+ was used meaning associated with (target and action in case 4).&- was used to show something excluded (action in case 5). Sometimes pathologic conditions were actual targets of actions. However they were not listed in target axis in the content model. The target in the target axis was put on the left side of ] to keep the original model, and the pathologic condition was put on the right side (target in case 6). Some actions required more than two targets as in sentence structure with indirect and direct objectives. Target as a role of direct object was located on the left side of >, and the other one as a role of indirect object was located on the right side (target in case 7). The model was successfully applied to the rest of the data. When it was applied to the obstetrics and gynecology data, it could gather the procedures labels with similar properties in a group and make rearrangement of the hierarchy easier as shown in Figure 2.We are currently waiting for the validation by doctors which will be presented during the conference.

5. Conclusion

On the whole the ICHI semantic model was able to represent most of the ICD 9CM Volume 3 and the specific Korean coding systems labels. Some difficulties still need to be overcome: to find the Action value, some extensions are needed for the number of accepted Targets and Actions and Pathology as a Target. This work is a case study showing how the ICHI international initiative can support the harmonization of national health interventions coding systems starting from the unofficial standard of ICD 9 CM Volume3.

References

[1] Rodrigues J-M, Kumar A, Bousquet C, Trombert B. Standards and Biomedical Terminologies: The CEN TC 251 and ISO TC 215 Categorial Structures. A Step towards increased interoperability. In: Andersen SK, et al. (Eds.) MIE 2008 Proc. IOS Press, 2008; pp. 735-740. [2] Madden R, Zaiss A, Thorsen G, Lewalle P, Rodrigues J-M, Weber S, Ustun, B: World Health Organization Family of International Classifications: Developing the International Classification of Health Interventions: Background, Need and Structure, WHO 2008. [3] Weber S, Rodrigues J-M, Madden R, Pickett D, Zaiss A, ten Napel H, Moskal L, Bartz C, Virtanen M. The ICHI content model, WHO 2009. [4] Madden R. World Health Organization Family of International Classifications: ICHI project plan WHO 2010. [5] Trombert Paviot B, Madden R, Zaiss A, Bousquet C, Kumar A, Rodrigues JM. Towards the International Classification of Health Interventions (ICHI). Step 2.Populating the ICHI content model with existing coding systems. In Proceedings of PCSI International Munich 2010.