The ICD-9/10/11 are released by the WHO and used worldwide to support applications including health insurance classification. However, different countries develop their own modifications of the ICD system and these versions are often incompatible. In addition, the semantic relations among ICD disease terms are unclear, and how these terms are related to other entities such as anatomic entities are not defined. To address these issues, we developed an ICD ontology (ICDO) to logically represent ICD terms and their relations with anatomic entities, qualities, etc. Different from other disease ontologies, all ICD diseases are defined disease processes in ICDO. The current ICDO focuses on English and Chinese representation. As a use case, we used ICDO to integrate ICD related data from 33 regions in Jiangsu province in China. Our strategy was able to identify and standardize local ICD versions in these regions.
The International Classification of Diseases (ICD), maintained
by the World Health Organization (WHO), is the international
standard for reporting diseases and health conditions. It is the
diagnostic classification standard for all clinical and research
purposes. ICD defines diseases, disorders, injuries and other
related health conditions in the biomedical and clinical domains
in a comprehensive and hierarchical fashion. The ICD has been
continuously revised and published in a series of editions to
reflect advances in health and medical science over time (
Many countries have adopted the ICD standard and developed
their own modified versions. For example, there are the USA
version of ICD-10-CM (
GB/T 14396-2016
Guangdong ICD-10 ( 2017）
Ji’nan city,Shandong province ICD10 (For Health
information data sharing and exchange )
Note: This source of this table comes from a survey by OAMAHA:http://www.sohu.com/a/302897591_324186, 2019
03-21, which is translated by author.
The ICD is used as the controlled terminology of diseases in the
medical information platform in most healthcare
administrations. There are many application systems that exist
in hospitals, such as: HIS (health information systems) (
Copyright © 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
On the other hand, both ICD codes and Diagnosis-related
Groups (DRGs) are a major method for medical insurance
control and the DRGs is dependent on the correctness of ICD
(
The semantic mapping among databases generated under two
different coding systems (e.g., ICD10 and ICD11) is very
difficult and generally requires manual intervention. The National
Institutes of Health (NIH) refer such difficulty to the
phenomenon of ‘data wrangling’ encompassing activities that make data
more usable by changing their forms but not their meanings (
Although great efforts have been made on this area, the obstacle still exists. The ICD terminology is composed of a code/value pair. Each ICD standard code corresponds to a unique disease name as a value. However, in reality, there are often multiple synonyms expressed for one disease in the natural language.
For example, the ICD 11 code AA0Z has value of Infectious diseases of external ear, unspecified; the GB/T14396-2016 code H60.001 has value of 外耳疖 (external ear furuncle); the ICD 10 code H60.5 has value of acute otitis externa, noninfective. Due to the existence of polysemy in natural language (especially in Chinese), the code-value mapping often encounters ambiguity after using Extraction-Transfer-Load (ETL) tool for data integration, and results in improper matching. Particularly in China, these problems are mainly due to the different local ICD versions with private extensions to certain ICD terms.
These modifications are made according to the internal clinical needs coming from different medical units. The large discrepancy among different versions might cause many problems, such as the appearance of the large amount of data with different values but the same code, or the same value with different codes. This also affects the accuracy of ICD-based DRG grouping, the accuracy of Medicare payments as well as the statistics accuracy of death causes.
In addition to the ICD, there are many disease discription
models being developed and used. Hadzic et al. classify disease
into four dimensions: (i) generic disease types; (ii) phenotypes
that are mainly based on observations to describe the various
symptoms of the disease; (iii) etiology that is a strictly scientific
basis of pathogenic factors, mainly including two categories
genetic factors and environmental factors; (iv) treatment that is
a possible effective measures against a particular disease (
These four dimensions together can describe the overall
knowledge of a disease field. Fang et al. learned from the
classification of SNOMED (
Ontology is likely the best approach to solve the issue of
semantic mapping among different databases and terminology
systems. A formal biomedical ontology is a set of computer and
human-interpretable terms that represent entities and relations
among the entities in a biomedical domain. Ontologies have
emerged to be critical to biomedical and clinical data
standardization, management, integration, and analysis. Two
different databases or terminologies may be formed based on
different organizational principles and are unlikely or difficult
to form an agreement about what each piece of information
refers to and how they can be aligned. The inability to achieving
interoperability can severely compromise the goals of
information integration and aggregation. Such issue is difficult
to solve internally or among the two databases (
Many disease-related ontologies exist, including Human
Disease Ontology (DOID) (
To find a semantic mapping method between different ICD
versions, here we report the development of an ICD ontology
(ICDO) to address the issues of database interoperability and
data integration as listed above. Given that ICD is mainly
applicable to statistical analysis and disease grouping for
healthcare insurance, we present in this paper our disease
design pattern that combines the advantages of the above
disease description models. Our disease design pattern in ICDO
is based on the understanding that the disease in ICD is a human
pathological process that realizes disease disposition. Such
process is composed of a group of entities, which has reversible
decomposition. These entity are ‘anatomical structure’,
‘pathological anatomical entity’, ‘etiology’, ‘quality’ and
‘syndrome’. Therefore, all the ICD terms are defined as
subclasses of the ICDO “disease process” class, which is then
defined as a subclass of the imported OGMS term ‘pathological
bodily process’ (
Our ICDO development closely followed the WHO ICD 10/11
classification and principles. The ICDO development used the
eXtensive Ontology Development (XOD) strategy (
We focused our first stage ICDO development on the specific area of external ear diseases as a proof-of-concept. This early stage ICDO prototype includes all diseases related to external ear part in: (i) ICD11 under the class, “Disease of the ear and mastoid process”, coded from AA00 to AA6Z, (ii) ICD10 under the “external ear diseases”, and (iii) GB/T 14396-2016. The Protégé OWL editor (http://protege.stanford.edu) was used to visualize ICDO, add new ICDO terms, edit imported terms and merge imported ontologies. ICDO-specific terms were generated using new ICDO identifiers with the prefix “ICDO_” followed by 7-digit auto incremented numbers. The Hermit reasoner was used for consistency checking and reasoning (http://hermit-reasoner.com/).
ICDO is expressed using the W3C standard Web On-tology Language (OWL2) (http://www.w3.org/TR/owl-guide/). The current ICDO source code is openly available at GitHub: http://github.com/icdo/ICDO.
The ICDO ontology is deposited in the NCBO BioPortal
website: https://bioportal.bioontology.org/ontologies/ICDO,
as well as the ontology repository website Ontobee (
The health systems in several regions of the Jiangsu Province in China used different modified versions of ICD10. To standardize the coding systems from these regional health information platforms, we identified three regional ICD10 modification coding systems, and used ICDO to model and standardize these coding systems.
Description Logic (DL) query was used to query the knowledge built in ICDO. The DL query function in the Protégé-OWL editor was used for the implementation.
First we performed a survey of how the term “disease” is
defined in different ontologies and dictionaries (Table 1). It is
clear that the nature of disease is defined differntly. In four
ontologies including DOID, OGMS (
In OGMS, there are two disease-related terms ‘disease course’, and ‘pathological bodily process’. The term ‘disease course is defined as “The totality of all processes through which a given disease instance is realized”. However, it is unclear what the “all processes” in the definition means. It is possibly that some of the processes are not directly related to disease. The OGMS term ‘pathological bodily process’ is defined as “A bodily process that is clinically abnormal”. The diseases listed in ICDO have already happened, and are not an upcoming event. Given that the ICD is used primarily for post-disease recording and insurance filing purposes, we think that the disease in ICD is primarily meant to be a type of pathological bodily process; therefore, the disease in ICD can be better regarded as a “disease process” under OGMS ‘pathological bodily process’. In ICDO, based on the nature of ICD and its applications, we focus on the representation of disease processes instead. Therefore, the term ‘disease process’ becomes our major term, which is defined in ICDO as follows:
Disease process =def. a pathological bodily process that occurs in a specific anatomic location, realizes a disease disposition, has abnormal bodily phenotype, and results in a pathological anotomic entity.
Therefore, all the specific diseases in ICDO are all defined as disease processes, which are different from other disease description frameworks. As a result, ICDO represents all disease names from ICD11, ICD10, GB/T14396 as disease processes, often abbreviaetd with the suffix “DP” in ICDO term labels.
In this study, ICDO is mainly used to support data standardization among different ICD versions. ICDO aims to standardize clinical data from international multi-center and also data generated under different ICD local and modified versions in China. To support the general interoperability goal, we have included ICD10 and ICD11 terms in both English and Chinese languages in the ICDO.
ICD 10/11 has different classification and principles in top level
design and therefore we closely have followed OBO to develop
ICDO top level hierarchy. Extending from the formal definition
and classificaiton of this ICDO “disease process” term, we
generated an upper level ICDO hierarchical structure (Fig. 1).
ICDO reused many terms from existing ontologies such as the
BFO (
In ICDO, a disease process was composed of four major elements: etiology entity, quality, anatomical stucture and pathological anatomical entity. The disease pattern of ICDO was shown in Fig. 2:
Fig 2. ICDO disease process pattern We represented different diease processes following the disease process pattern (Fig. 2). For example, the ICD term ‘granuloma of external ear canal’ is defined in ICDO as “granuloma of external ear canal DP”, which is a granuloma disease process that “occurs in” some “external ear canal” and “has disease output” some “granuloma”. The “necrotic external ear otitis” is an otitis disease process “occurs in” some “external ear” and “has quality” some “necrotic”. Note that the ‘has quality’ is indeed a shortcut relation where the quality is not the quality of the process per se. Instead, the quality is the quality of the anatomical entity of the patient.
ICDO also has different design strategy for diseases compared to DOID and MONDO. In general, DOID and MONDO do not discomposed a disease term into different entity components such as etiology entity, quality and anatomical structure. We paid a lot of attention to each of these issues and developed our specific strategeis. In addition, due to the nature of ICD usage in clinical disease classificaiton and insurance filing, we have designed many special design patterns for the ICDO generation.
Some of the special design patterns, together with the approaches proposed in ICDO, are described in next session.
Fig. 2. Disease process modeling in ICDO. In this example, the term ‘obstructive keratosis of external ear DP’ is represented using the design pattern, including disease quality, occurs in location, and disease material output.
Besides general disease classifications, ICD includes many special terms such as “classified elsewhere”, “other specified” and “unspecified”. ICDO has implemented special strategies to handle the mentioned special terms.
The first special terms, “classified elsewhere”, were treated as obsoleted terms in ICDO. We believe that the disease classification must be clear and consistent among various disease categories. The definition of “classified elsewhere” is confusing because there is no obvious and proper disease category for “elsewhere”. To ensure the classification integrity, a disease term can be classified under multiple disease categories based on varies definitions and applications, but it should not be classified as an undefined category, “elsewhere”. To balance the mapping process among various ICD versions and proper handling of the undefined category, we added all the ‘disease classified elsewhere’ terms in ICDO but made them as obsoleted terms in the ontology.
There are also many ICD terms labeled as “other specified”. Logically speaking, all ICD terms should be classified into specific classes and there should not exist any ‘other’ class. We consider this type of “other specified” terms class as logical error and put all the terms under this class into their parent class. In other words, we generated an ICDO term “other specified” and put it under the obsolete to support mapping among to existing ICD versions. To ensure the continuity of the various versions of the ICD in the conversion adaptation process, in the data adaptation process, this obsolete class term may still participate in the operation to ensure the accuracy of data mapping. Many ICD terms are labeled as “unspecified”. For their corresponding classes, we have been able to determine the parent classes. However, due to the limitations of current definitions and the lack of knowledge, these “unspecified” terms have not given any specific description as of now, and can be mapped to their parent terms in ICDO.
The ICD10/11 has the Extension Codes used to support clinical treatment, such as the organ laterality in different ICD versions and the “special anatomy” in ICD11. The laterality commonly found in various ICD versions includes “left”, “right”, “bilateral”, “unilateral, unspecified”, “unspecified laterality”. Additionally, ICD11 introduced a term “special anatomy”, which includes anatomical synonyms and possible anatomical structure of disease. We built the laterality as the quality of the disease in ICDO, and adopted the synonym of the specific anatomy as a synonym label in annotation in anatomical structure if UBERON have not included the synonym. Other extension codes exist such as distribution and regional. For example, abscess of right external ear DP can be defined to have axiom assertion of: “abscess of external ear DP” and (“occurs in anatomical side” some “right side of anatomical entity”) Such design properly handles the issue of anatomic literality. With the focus of disease process, ICDO also has a natural advantage of defining different disease stages, or the beginning, middle, and end of a disease process. Such a process aspect supports real life disease representation.
The development of ICDO starts from designing the disease pattern first, then decomposing the terms into different components of the semantic equivalence terms. Then we used the Ontobee annotator (http://www.ontobee.org/annotate) to discompose these disease terms into different components and tag the components according to our disease design pattern (Fig. 2). Finally, we established relationships between the components of the disease terms by creating objective properties and annotations following the disease desgin pattern.
This entire process involves the decomposition of a disease name and then establish the logical relations among the components. (Fig. 3).
In order to illustrate the process, we extracted terms from three local ICD10 versions genereated by two districts and one city (Pukou district, Liuhe district and Jiangyin city) in Jiangsu province, China, and performed the mapping. The China administrative departments from different regions often publish and adopt their own ICD versions. Therefore, the harmonization of the ICD in China is a complex and difficult issue due to varies local versions and custom modifications. As shown in Table 1, there are 14 different ICD local and modified versions in China. Among them, GB/T-14396-2016 is a Chinese version of ICD10 modification required by the Chinese government since February 2017. Some local versions listed in Table 1 are used for clinical purposes and the others used for adminstrative statistics purposes. For example, in Jiangsu province in China, we identified more than 30 regional datasets but they used different locally modifed ICD versions. Even though the Jiangsu province has the most advanced health informatics system among all the provinces in China, many incorrect code-value pairs existed in these local coding systems and need to be mapped to the GB/T-14396-2016 coding system.
(i) Same code but different values: For example, in the local Liuhe district coding system, the two Chinese disease term “坏 死 性 外 耳 炎 ” and “ 恶 性 外 耳 炎 ” (English translation: “necrotic external ear otitis” and “malignant otitis externa” respectively) have the same code H60.200. However, the code H60.200 corresponds to the “malignant otitis externa by ICD10 or GB/T14396-2016, and the term “necrotic external ear otitis” should be coded as H60.900 (“otitis, externa”).
(ii) Same values but different codes: For exmaple , in the local Jiangyin city coding system, the Chinese disease term “后 天性外耳畸形” (English translation: “acquired deformity of ear externa”) has two codes: H61.303 and H61.101. However, according to ICD10 or GB/T14396-2016, the correct code of this disease name should be H61.101. The code H61.303 even does not exist in ICD10 or GB/T14396-2016.
These two types of errors shown above widely exist in the local Chinese ICD modified versions. For example, even for the same external ear branch, there are 36 errors in Pukou district, 4 errors in Liuhe disctrict and 14 errors in Jiangyin city local coding systems. Note that there are only 22 code-value pairs in ICD10 and 41 code-value pairs in GB/T14396-2016 for the external ear-related diseases. Considering the total number of over 20,000 terms in ICD10 and 14 different local ICD10 versions in China (Table 2), it is a huge effort to manually correct these code-value pair errors. the pair errors have become a major issue to support data integration and systematic statistical data analysis in China.
In this study, we developed an ICDO-based semantic disease name mapping algorithm (ISDNA), as shown in Fig. 4, with the aim to solve the disease name mapping issue as illustrated in the use cases of Jiangsu Province, China. First, the ISDNA algorithm first accepted some disease names in a specific langugae (e.g., English, Chinese) as input. The input names were then decomposed via natual language process (NLP) to different components using Ontobee Annotator. For example, a disease name could be broken down into the anatomic entity term as the location of the disease process, quality of the patient or the anatomic bodily entity of the patient, and abnormal pathological entity as the output of the disease. These identified components were then mapped to their corresponding ontology terms and IDs. Based on the axioms defined in ICDO, we can use ontology reasoners to automatically infer the ICDO codes that the input disease name belongs to (Fig. 5). Our ISDNA algorithm is able to identify exact matches that perfectly mapped or semantically inferred the parent terms of the matched disease names from the Pukou district, Liuhe district and Jiangyin city in Jiangsu province, China to the corresponding ICDO terms. Next we will provide two examples to illustrate the features and performance of the ISDNA algorithm. Fig. 5. provides the first example of how we can infer a specific name to a perfectly matched ICDO term. Specifically, in this use case, “cellulitis of external ear” (Chinese name: “外耳蜂窝 织 炎 ”) is an input disease term. It was first split into two components: “cellulitis anatomic entity” and “external ear”.
Given the nature of these two terms, the following two axioms could be assigned: ‘occurs in’ some ‘external ear’ ‘has disease output’ some ‘cellulitis anatomic entity’ Based on these two axioms, the ICDO reasoner was able to infer “cellulitis of external ear” as an exact match to the ICDO “cellulitis of external ear DP” (Chinese name “外耳蜂窝织 炎”) term with the GB/T14396 code H60.100, ICD10 code H60.1, and ICD11code AA01. Then we can select one of the code from them according to our needs.
Fig. 6. provides another example that has the input disease name “Pinna defect after burn” (Chinese name: “烧伤后耳廓 缺损”), a term from the Pukou district local coding system in our use case. Similarly, our algorithm started with spliting the long disease name to three components: “pinna”, “defect”, and “after burn”, which could then be mapped to their corresponding ontology terms in ICDO. Note that the term “after burn” is an defined as an synonym of the “acquired after burn”, which is a subclass of the quality “acquired”. The term “Pinna defect after burn” is not included in either ICD or ICDO.
But for demonstrative purpose, the term “Pinna defect after burn” was introducted (Fig. 6) to simulate the situation that there is no perfect matching. After direct mapping, there was no exact match for this disease name. However, after running the Hermit reasoner available in the Protégé OWL editor, we were able to infer this disease name to be subclass of the ICDO “acquired deformity of pinna DP” term (Chinese name: “获得 性耳廓畸形”) with ICD11 code AA41. Given that there was no exact match for “Pinna defect after burn”, the ICDO “acquired deformity of pinna DP” term was defined as the preferred semantically matched ICDO term for the input disease name.
Fig.5. ISDNA inferred terms exact mapping include different codes and languages come from different codeing systems.The candidate term “cellulitis of external ear”will be discomposed into components as “cellulitis DP” and “external ear” after NLP first.Then they are mapped to respect terms in ICDO according to the dimensions disgned by disease pattern as “cellulitis DP” and “external ear”.Finally inferred to “cellulitis of external ear DP” by axioms in ICDO .User can selects different ICD code by application requirement.
Note that in the above two examples, the NLP process was preformed manually. In the future, we plan to develop an automatic NLP process to achieve the same NLP results, which is not within the scope of current study.
In this case, we demonstrate how to use the Description Logic (DL) query in the Protege-OWL editor to identify from ICDO specific diseases that occurs in the external ear canal, or called external acoustic meatus. Basically, this DL query identified those diseases that meet this axiom requirement: “occurs in” some “external acoustic meatus” The “external acoustic meatus” is the formal anantomical structure term in UBERON and has synonam “external ear canal” in ICDO.
As shown in Fig. 7, a total of 14 diseases were identified, including 4 testing terms in Chinese. This example shows that ICDO is able to serve as a platform and knowledge system for computational programs like DL query to perform semantic query and analysis.
Fig. 6. ISDNA mapping of a candidate term to ICD standard term and code. In this example, the term ‘Pinna defect after burn’ (Chinese name: “烧伤后耳廓缺损”) was used as the input. The term was decomposed into three components “after burn”, “pinna”, and “defect”, where were then mapped to ‘acquired from burn’, pinna, and defect in ICDO, respectively. These three ICDO terms provide the quality, location of the disease process, and the output of the disease process term. Based on the axiom definition, our ontology reasoner was able to match this name to ‘acquired deformity of pinna DP’ (AA41 in ICD11). Note that this term is not an exact match.
Fig. 7. DL query of ICDO looking for all disease “occurs in external acoustic meatus”. This query was performed using DL query in Protege-OWL editor 5.2 (http://protege.stanford.edu/).
In this manuscript, we presented our development of the ICDO ontology with the aim to standardize ICD disease records and support health record integration and analysis. We also proposed and tested a semantic analysis based on ICDO using the function of reasoner. It realized the interpretation of terms at the semantic level by reasoner between entities by axioms. ICDO improves the mapping accuracy, supports exact and semantically preferred mapping. and provides a useful application in terms of the standardization of heterogeneous data between different ICD versions.
Our use case focused on the different ICD10 local versions used in some local adminstrative healthcare information plateform in Jiangsu province China. Not every disease has a clear physical product "disease output" when entering clinical observation. For example, inflammation is an immune with multiple symptoms and are sometimes difficult to fully express in natural language. However, the physical entity of this inflammation is clearly present as specific anatomy structure in ontology. In ICDO we defined the output of inflammation process with “inflammatory anatomical entity” and asserted axiom in the form of “physical pathological object” “occurs in” some “anatomical structure” “caused by” “inflammation process”.
In clinical practice, our disease pattern can cover most disease types. Particularly there is a class of diseases which does not have pathological abnormalities in specific anatomical structures but have systemic symptoms. We designed a special dimension “syndrome” in ICDO disease pattern for this class of diseases. In the current stage of ICDO with the focus of external ear disease, there is no syndrome included. However, this situation will be carefully examined and appropriately handled in the future when we extend the ICDO to fullly cover all disease in ICD10 or ICD 11.
While many ICD terms can be clearly defined as disease processes, there may be many concerns in terms of using disease process for other scenarios. For example, the disease process may not be able to represent a disease output such as the size or mass of a tumor like external ear canal tumor. However, in this case, we can semantically define a disease process like ‘external ear canal tumor disease process’ that ‘has output’ of some tumor that ‘has size’ or ‘has mass’ of some specific values. Using this strategy, we can semantically link the disease process to the physical tumor (or other anatomic entity) and its qualities like size or mass. Another concern is that a disease process may be diagnosed before the result of the process becomes manifested. Note that ICD is typically used to represent the health outputs rather than unidentified or undiagnosed health issues. The fact that a disease is not diagnosed does not mean that the disease process does not occur. If the disease is not diagnosed, we may not be able to use the disease process term; however, it may not be necessary to use according to the ICD guidelines.
Ontology is clearly a very good tool for solving the problem of semantic mapping between different ICD versions. ICDO will improve the usability and interoperability among various ICD systems. ICDO can also be used for data standardization and analysis of international multi-center clinical trials between different languages in different counties, data normalization processing before DGRs grouping, data normalization and in hospital internal information systems, and data standardization for regional health information platform. The disease design pattern in ICDO can provide effective contributions to the medical data mining and retrospective researches.
We appreciate the discussion and editorial revision by Ms. Meng Liu.
LW and YH are co-corresponding authors. Their emails addresses are wanlingeric@qq.com and yongqunh@med.umich.edu.