Language of Czech Medical Reports and Classification Systems in Medicine, En 58-65

en58 P. Přečková: Language of Czech Medical Reports and Classification Systems in Medicine, en 58-65

Language of Czech Medical Reports and Classification Systems in Medicine

P. Přečková1 1Centre of Biomedical Informatics, Institute of Computer Science AS CR, Prague, Czech Republic Supervisor: Prof. RNDr. Jana Zvárová, DrSc.

Summary interoperability, semantic interoperability, attributes to various classification systems The objective of the paper is to compare cardiology, atherosclerosis is a very important step for semantic Czech medical reports written in a free text interoperability among these various and by means of a software application; to 1. Introduction s y s t e m s . E H R s a n d s e m a n t i c analyze the usability of international Determination, denomination, and interoperability are very topical issues and classification systems in the Czech classification of medical terms are not they are discussed in many papers [3], [4], healthcare environment. The analysis of optimal. The proof is that for one term we [5], [6]. Semantic interoperability based on medical reports was based on the can often meet with more than ten the Czech language has been studied in attributes of the Minimal Data Model for synonyms. This problem has intensified [7], [8], [9], [10], [11]. Cardiology (MDMC). We have used with an introduction of a computer medical reports written in a free text and technology to healthcare. Using compu- Clinical data from electronic health records medical reports from the ADAMEK ters means higher uniqueness of data has traditionally contained a small software application where data are stored feeding, of term definitions, their precise proportion of fixed field data (often in a structured way. For our work SNOMED denomination, etc., thereby the significant obtained from pick lists) and larger CT and ICD-10 have been used. We have drawback becomes more noticeable. quantities of a free text [12]. In our paper focused on the language of Czech medical we will discuss both of these types of data. reports and the application of Generally, in the scientific terminology it is aforementioned international classification more advantageous to use only one 2.Coding and classification systems systems in MDMC. We have compared expression for one term. Computers are Coding systems limit the variability of how well attributes of MDMC are recorded able to learn synonyms but it enlarges expression. Only the approved terms and in textual medical reports and in medical dictionary databases and the number of their phrases can be used according to reports recorded structurally by means of necessary operations grows. Moreover, strictly given rules. Formal codes are the ADAMEK software application. We synonymy in the scientific terminology usually used instead of the approved have made the language analysis of the leads to inaccuracy and misunder- terms. In many cases it is useful if coding Czech textual medical reports. We standing. In current medicine we can meet systems show also not approved terms, compared how MDMC attributes are with many synonyms for one single which are often used as synonyms for recorded in the ADAMEK application and disease. That was the reason why coding approved terms. in medical reports written in a free text. To systems providing codes for any medical conclude, using a free text in medical findings have arisen. Let us show you some of the most reports is very inhomogeneous and not widespread international classification standardized. The standardized Nowadays, there is a big boom in the systems. terminology would bring benefits to development of electronic health records physicians, patients, administrators, (EHR). There is a general agreement that 2.1 ICD International Classification software developers and payers. It would electronic, computer-based medical of Diseases and Related Health help healthcare providers that it could records have the potential to improve the Problems provide complete and easily accessible quality of medial care [1]. Within the The foundation of the International information that belongs to the process of concept of the EHR, the patient is Classification of Diseases was laid by healthcare and it would result in better care understood as an active partner who is William Farr in the year 1855. The World of patients. The use of international accessing, adding and managing health- Health Organization took it over in the year classification systems is a necessary first related data [2]. 1948. At that time it was its 6th revision. step to enable interoperability of Since 1994 the 10th revision of ICD is in heterogeneous electronic health records. Safe and appropriate clinical information use and it contains 22 chapters. ICD has exchange among various electronic health become an international standard for Keywords: terminology, synonyms, records is essential for continuity of a classification of diseases and for many classification systems, thesaurus, patients' care in various times, at various epidemiological and management needs nomenclature, electronic health record, places and in various healthcare providers. in healthcare. Mapping of electronic health record

These include a general situation of health scientific literature or in the arising fields of (COSTART); Diseases Database; in different population groups and research. They define these terms in the DXplain; Gene Ontology (GO); Healthcare monitoring of the incidence and frame of the contents of the existing Common Procedure Coding System prevalence of various diseases and other vocabulary and they recommend their ( H C P C S ) ; H o m e H e a l t h C a r e health problems in relation to other adding to the MeSH vocabulary. There Classification (HHCC); Health Level variables. It is used to classify diseases exists also the Czech translation of MeSH. Seven Vocabulary (HL7); Master Drug and other health problems that are Data Base (MDDB); Medical Dictionary for recorded in many types of health records, 2.4 LOINC Regulatory Activities Terminology including death certificates and hospital Logical Observations Identifiers Names (MedDRA); Multum MediSource Lexicon records [13]. ICD is available in six official and Codes (LOINC) [18] is a clinical (MMSL); NANDA nursing diagnoses; NCBI languages of WHO and in other 36 terminology, which is important for Taxonomy and many others. languages, including Czech. laboratory tests and laboratory results. In the year 1999 the HL7 organization 3.1 Conversion tools 2.2 SNOMED CT accepted LOINC as a preferred coding The increasing number of classification SNOMED Clinical Terms originated from system for names of laboratory tests and systems and nomenclatures requires two terminologies: SNOMED RT and clinical observations designing of various conversion tools for Clinical Terms Version 3 (Read Codes transfer among main classification CTV3). SNOMED CT represents the 2.5 ICD-O systems and for recording of relations Systematized Nomenclature of Medicine The ICD-O classification system [19] is an among terms in these systems. Extensive Reference Terminology developed by the extension of the International classification ontologies and semantic networks are College of American Pathologists. It serves of diseases for oncology coding. It is a four- modeled for information transfer among as a common reference terminology for dimensional system. The dimensions are various databases. Metathesauri are gathering and acquiring health data appointed to classify morphological kinds designed to monitor and connect recorded by organizations or individuals. of tumors. The third version of ICD-O is information from various heterogeneous The Clinical Terms Version 3 was used nowadays. sources. developed by the United Kingdom's National Health Service in the year 1980 as 2.6 TNM The Unified Medical Language System a mechanism for storing structured The TNM classification system [20] is a (UMLS) [21] was initiated in the year 1986 information on primary care in Great clinical classification of malignant tumors in the National Library of Medicine in the Britain. These two terminologies united in used for comparison of therapeutic USA. UMLS knowledge sources are the year 1999 and a highly complex studies. The TNM system is based on the universal. It means they are not optimized terminology SNOMED CT arose [14], [15], assessment of three components: T - the for individual applications. It is an [16]. Nowadays we can meet with extent of the primary tumor; N - the intelligent automated system, which American, British, Spanish, and German absence or presence and extent of “understands” biomedical terms and their versions of SNOMED CT. regional lymph node metastasis and M - relations and it uses this understanding for the absence or presence of distant reading and organization of information 2.3 MeSH metastasis. It proceeds from the from machine processed sources. Its aim Medical Subject Headings [17] is a knowledge that, for the disease prognosis, is to compensate terminological and vocabulary controlled by the National the localization and spread of a tumor is the coding differences of these non- Library of Medicine (NLM). It is composed most important. homogeneous systems and also language of terms, which denominate keywords varieties of users. It is a multilingual hierarchically and this hierarchy helps with 2.7 Other Classification Systems thesaurus of classification systems on a searching on various levels of specificity. Currently, there are more than one high-capacity medium, which enables to Keywords are arranged not only hundred of various classification systems. transfer coded terms among various alphabetically but also hierarchically. NLM These are for example AI/RHEUM; classification systems. uses MeSH for indexing of papers from Alternative Billing Concepts (ABC); Alcohol world best biomedical journals for the and Other Drug Thesaurus (AOD); Beth UMLS is based on three knowledge MEDLINE/PubMED database. MeSH is Israel Vocabulary; Canonical Clinical sources: Metathesaurus, Semantic used also for a database cataloguing Problem Statement System (CCPSS); Network, and SPECIALIST Lexicon. The books, documents, and audiovisual Current Dental Terminology (CDT); DSM Semantic Network contains information materials. Each bibliographical reference (Diagnostic and Statistical Manual of about semantic types and their relations. is connected with a class of terms in the Mental Disorder); Medical Entities The SPECIALIST Lexicon records MeSH classification system. Searching Dictionary (MED); Current Procedural syntactic, morphologic, and orthographic inquiries use also the MeSH vocabulary to Terminology (CPT); International information of each word or a term. find papers with required topics. The MeSH Classification of Primary Care (ICPC); vocabulary is updated continuously and it McMaster University Epidemiology Terms; is also controlled by specialists creating it. CRISP Thesaurus; Coding Symbols for a They collect new terms appearing in Thesaurus of Adverse Reaction Terms

The UMLS Metathesaurus is an extensive, analyzed medical reports have been 5. The language of Czech medical multi-purpose, and multilingual database. anonymous and therefore without any reports and the application of It contains information about biomedical, administrative data. international classification systems healthcare and their relative terms, their MDMC consists of eight groups of in MDMC various expressions and relations among attributes. After the administrative part 5.1 The Czech language them. Its main aim is to connect alternative there is a family history part consisting of The Czech language belongs to the expressions of the same terms and to information about mother, father and western group of Slavic languages. As a identify useful relations among various siblings. The next part is the social history Slavic language Czech belongs to the terms. If thesauri use the same and addiction focusing on the marital eastern, or satem, division of Indo- expressions for different terms, then both status, physical activities, mental stress, European languages. The Czech meanings are present in the Meta- levels of smoking and alcohol consumption language separated itself from other Slavic thesaurus and we can also see which rates. One part of MDMC is devoted to languages by a number of changes, most meaning is used in which thesaurus. If the allergies, mainly to drug allergies. The of which took place in the 10th through 16th same term is used in different hierarchical personal history part detects the presence centuries. By the end of the 15th century contexts in various thesauri, then the of diabetes mellitus, there is observed the Czech language had virtually lost the Metathesaurus keeps all these whether the patient suffered from a stroke, dual number and two of the Slavic past hierarchies. The Metathesurus does not whether he/she is treated with an ischemic tenses. On the other hand, the role of the give one consistent view, but it keeps many disease of periphery arteries, there are verbal aspect had grown more significant views, which are present in source attributes related to aortic aneurysm, other and the number of declensions had thesauri. relevant diseases and menopause in increased. At the beginning of the 15th women. In the part of MDMC called century the religious reformer Jan Hus 4. Minimal data model Current difficulties of a possible (John Huss) devised a diacritical writing for cardiology cardiological origin physicians are focused system, placing diacritical marks over Since the year 2000 our department has on the shortness of breath, chest pain, some Latin letters to distinguish the Czech been involved in the research of the joint palpitations, swellings, syncope, cough, palatal/palatalized consonants (č, ď, ň, ř, š, workplace consisting of two universities, hemoptysis, and claudication. Another ť, ž) and long vowels (á, é, í, ó, ú, ý). In the two hospitals and the Institute of Computer part of MDMC determines what kind of a 16th century the letter “ů”, indicating the Science AS CR. This joint workplace is treatment the patient undergoes, what type long “u”, was added. The only diagraph called the EuroMISE Centre and it runs of a diet is prescribed and which surviving in modern Czech is “ch”. also two outpatient departments of medications he/she uses. In the part of the preventive cardiology. In the year 2002 the physical examination, the patient's weight, The Czech language has about 10 million Minimal Data Model for Cardiology height, body temperature, BMI, WHR, speakers in the Czech Republic and about (MDMC) was developed [22], [23]. MDMC blood pressure, pulse and breathing rates 200 000 speakers in other countries. is a set of approximately 150 attributes, and pathological finding are determined. These are mostly emigrants and children their mutual relations, integrity restrictions, Laboratory testing is focused on blood of emigrants who left the country in sizable units, etc. Prominent professionals in the glucose, uric acid, total cholesterol, HDL- migrations around World War I, World War field of Czech cardiology agreed on these cholesterol, LDL-cholesterol and II, and the year 1948 and 1968. Many attributes as on the basic data necessary triaglycerols. The last part of the MDMK is Czech speakers can be found especially in for an examination of a patient in focused on attributes related to ECG. The Austria (mostly in Vienna), Poland, cardiology. beat frequency, the average PQ and QRS Germany, Ukraine, Croatia, in Western intervals and the full description of ECG is Romania, in Australia, and Canada. As cardiology is a very extensive field described there. However, the largest group of Czech MDMC is limited to atherosclerotic speakers outside the Czech Republic lives cardiovascular diseases. The aim of MDMC has become a basis for the in the United States of America, in cities MDMC has been to form a minimal set of ADAMEK software application where data like New York, Chicago and in a number of attributes, which are needed to be are stored in a structured way. After its rural communities in Texas, Wisconsin, observed in all patients from the completion the data collection started in Minnesota, and Nebraska. cardiological point of view so a patient March 2002. The data were and still are could be ranked into ill persons or persons collected in two outpatient departments of The Czech language belongs to the free in risk from the perspective of preventive cardiology of the EuroMISE constituent order languages [24]. cardiovascular diseases. Centre. To April 1st, 2010 there were data about 1189 patients. 5.2 The language analysis of the Czech Necessary administrative attributes are medical reports written in a free text one part of MDMC. Nevertheless, they are The style of writing medical reports is not not included in our analysis as the standardized. We can find differences not

EJBI – Volume 6 (2010), Issue 1 © 2010 EuroMISE s.r.o. P. Přečková: Language of Czech Medical Reports and Classification Systems in Medicine, en 58-65 en61 only in medical reports from various one physician rounds one attribute to If we look at individual attributes, we can physicians but also individual medical integers, while another physician rounds see that only diastolic and systolic doctors write the same concepts in the same attribute with the precision of one pressures were recorded in all textual different forms. In the following part we will or two decimal numbers. Sometimes medical reports that have been analyzed. concentrate on linguistic and lexical numerical values are presented as ranges, In 96.30 % of the textual medical reports differences in Czech medical reports e.g. “70-80”. Often only an approximate medications that a patient uses or a written in a free text. indication is entered, e.g. “diastolic physician administers are recorded. Also pressure around 70”. Some attributes are the weight is recorded in 96.30 % of Diacritic: As we have formerly stated the not expressed in numbers but in words, medical reports. In contrast, the height is Czech language uses the diacritical writing e.g. “blood pressure is within the normal recorded only in 74.07 % of reports. In 27 system. As an example of diacritical letters range”. analyzed reports some of the MDMC let us mention for example letters “ě, č, ř, ž,” attributes have not appeared not even in and others. However, it is faster for Arabic and Roman numerals: There is one case. This includes the following physicians to write without these diacritical a difference in usage of Arabic and Roman attributes: aortic aneurysm, angina marks and use letters “e, c, r, z”. Such a text numerals. For example heart sounds may pectoris, ischemic disease of lower limbs, i s f o r C z e c h n a t i v e s p e a k e r s be found in both ways “heart sounds 2” and when DM was discovered, silent ischemia, understandable but it is difficult for “heart sounds II”. body temperature and others. Drug allergy computational processing. is mentioned in 22.22 % of reports, Synonyms: The Czech language is very whether a patient drinks or does not drink Typing errors: Typing errors represent rich in synonyms and they are highly used an alcohol in 51.85 % reports, chest pain in a bigger problem and they are very also in medical reports. 37.04 %. The total psychical stress is frequent. The text is then very hardly recorded in 11.11 % of reports, physical usable for computational processing. Orthography: Some physicians use load in a job in 11.11 % of reports, total a newer version of spelling, the others the cholesterol in 70.37 %, drinking of a black Spaces: A similar problem is spaces older one. coffee in 22.22 %, various other omitting between words, which results in examinations in 62.96 %. The breathing merging of two words in one and this is Time data: Recording of time is not frequency has been found in 3.7 % of again unusable for computational standardized as well. In medical reports we reports, diabetes mellitus in 40.74 %, diet processing. Physicians vary also in using can meet the name of the month, e.g. in 59.26 %, glycaemia in 51.85 %, HDL spaces in front of units. We may find a form “February 2006” but also the month order, cholesterol in 66.67 %. The presence or with a space, e.g. “2.5 mg” but also the e.g. “2/2006”. absence of hypertension has been form without a space, e.g. “4mg”. The recorded in 70.37 %, left ventricular same situation is with attributes using a Drugs administering: In the medical hypertrophy in 11.11 %, myocardial slash. Some physicians use the variant reports there appear a lot of various ways infarction in 14.81 %, the average amount without a space, e.g. “80/min”, the others how to describe the time when a patient of cigarettes in a smoker in 51.85 % of use a space, e.g. “70 / min”. should administer a drug (e.g. 1-0-0 vs. 1 reports and so on. pill in the morning vs. 1 in the morning vs. Figure 0: For computational processing it 1x in the morning). We can say that while recording results of is difficult if some physicians use instead of examinations by means of a free text a lot the figure 0 the capital letter O. Attribute values: The same values of the of attributes is left unrecorded. There may attributes are often recorded in different be several reasons for that. Physicians do Abbreviations: As physicians are often ways. For example: diabetes mellitus can not have a strictly given skeleton according pressed of time they abbreviate words be found in a form of diabet, diabet., which they should proceed and it can while writing medical records. The problem diabetes mellitus of the 2nd type, diabetes happen that they may forget some here is that there exists not a single rule mellitus of II type on a diet, DM of the 2nd attributes. In software applications it is not how particular attributes should be type. possible because if physicians forget to fill abbreviated. Therefore we can often meet in the given attribute, the application will diversely abbreviated same words. We can These are not problems only of writing not let them to continue. Another reason find also examples that in one medical medical reports but the same orthographic why some attributes are not recorded may record written by one physician one errors may be found e.g. in web pages [25]. be the fact that physicians from the attribute is abbreviated several times and previous attributes know that the next each time differently. With abbreviated 5.3 The analysis of MDMC attributes attribute cannot be present and that is the words the problem of full stop omitting in medical reports written in a free text reason why they do not ask about it and behind the abbreviated word arises. The analysis of medical reports written in they do not record it. a free text was based on the attributes of Rounding-off: Another part where we can MDMC. Medical reports were anonymous find many discrepancies is connected with and therefore administrative data were not numerical values. We can meet with that possible to analyze.

But from the free text medical report we do recorded in 917 reports of the software not know whether these missing attributes application, which is 82.0 %, while in free Tab. 1. Percentage of recorded values of the have been checked or whether they have text medical reports we can meet with this selected MDMC attributes in 1118 ADAMEK been deduced by physicians on the basis attribute recorded only in 66.7 % of medical reports and in 27 free text medical of previous knowledge. analyzed reports. Both kinds of reports are reports.

the closest in weight that is recorded in the MDMC attribute ADAMEK free text free text medical

5.4 The analysis of MDMC attributes in ADAMEK application in 97.9 % of medical medical medical reports – the ADAMEK software application reports and in 96.3 % of free text medical reports reports 95% confidence

For this analysis 1118 medical reports from reports. Presence or absence of interval the outpatient department of preventive hypertension can be met in 95.3 % of the n=1118 n=27 lower limit upper cardiology of the EuroMISE Centre have ADAMEK reports, while only in 70.4 % of limit been used. free text reports. A big difference can be drug allergy 100.0 % 22.2 % 8.6 % 42.3 % found in left ventricular hypertrophy. It is aneurysm of aorta 100.0 % 0.0 % 0.0 % 12.8 % In all medical reports generated by recorded in all medical reports from the angina pectoris 100.0 % 0.0 % 0.0 % 12.8 % ADAMEK, in 100 %, these MDMC ADAMEK application but only in 11.1 % of chest pain 100.0 % 37.0 % 19.4 % 57.6 % attributes were recorded: drug allergy, free text medical reports. Another big level of stress 96.2 % 11.1% 2.4 % 29.2 % aneurysm of aorta, angina pectoris, chest difference is e.g. in the menopause total cholesterol 83.4 % 70.4 % 49.8 % 86.2 % pain, swelling of lower limbs, asthma, left attribute that is recorded in 96.9 % of diabetes mellitus 95.9 % 40.7 % 22.4 % 61.2 % ventricular hypertrophy, myocardial reports from the ADAMEK application but asthma 100.0 % 55.6 % 35.3 % 74.5 % infarction, other allergies, cough after ACE only in 7.4 % of free text medical reports. physical load in job 94.8 % 11.1 % 2.4 % 29.2 % inhibitors, claudication, silent myocardial The PQ interval is recorded by means of glycaemia 77.6 % 51.9 % 32.0 % 71.3 % ischemia, systolic pressure, type of DM the software application in 89.1 % and in HDL cholesterol 82.0 % 66.7 % 46.0 % 83.5 % treatment. Even from this enumeration it is 62.9 % in free text medical reports. weight 97.9 % 96.3 % 81.0 % 99.9 % clear that by means of the software Similarly the QRS interval is in ADAMEK hypertension 95.3 % 70.4 % 49.8 % 86.2 % application a higher number of attributes is medical reports recorded in 89.5 % and in myocardial infarction 100.0 % 14.8 % 4.2 % 33.7 % recorded. 66.7 % in free text medical reports. PQ interval 89.1 % 62.9 % 42.4 % 80.6 % other allergies 100.0 % 18.5 % 6.3 % 38.1 % If we compare how individual attributes are In Table 1 there is lucidly displayed the claudication 100.0 % 3.7 % 0.9 % 19.0 % recorded in free text medical reports of the percentage of recorded selected attributes smoker 96.5 % 66.7 % 46.0 % 83.5 % hospital and in ADAMEK medical reports, in MDMC in medical reports using the we reach the following results, see Table 1. ADAMEK application and in medical Drug allergy is recorded in free text reports written in a free text. Tab. 2. Selected attributes of MDMC coded by medical reports in 22.2 %, in the ADAMEK means of SNOMED CT. application in all, i.e. 100 % of reports. The 5.5 Attributes of MDMC coded by Attributes of MDMC SNOMED CT English equivalent answer whether a patient suffers or not means of SNOMED CT (in Czech) (Concept ID) from aneurysm of aorta has been recorded Table 2 shows several examples of MDMC alergie na léky Drug allergy (disorder) 416098002 in all reports from the ADAMEK application attributes to which the ConceptID from the Allergic reaction to drug (disorder) 416093006 but in none of textual medical reports. SNOMED classification system has been hypertenze Essential hypertension (disorder) 59621000 Questions about a chest pain have been allocated. The first prerequisite for this High blood pressure (&essential 194757006 recorded in all medical reports using the coding is the translation of the attributes to hypertension) ADAMEK application but only 37.0 % of the English language as there is not Essential hypertension NOS (disorder) 266228004 free text medical reports included remarks a Czech version, yet. ischemická choroba Ischemic heart disease (disorder) 414545008 on a chest pain. The level of stress has srdeční been recorded in 96.2 % of the ADAMEK 5.6 Attributes of MDMC coded by dušnost Asthma (disorder) 187687003 bolest na hrudi Dull chest pain (finding) 3368006 medical reports, in free text reports it was means of ICD-10 palpitace (Palpitations) or (awareness of 161965005 only in 11.1 %. The same percentage was As ICD-10 is one of a few international heartbeat) or (fluttering of heart) reached in the attribute of physical load in medical classification systems translated otoky Swelling or edema (finding) 248477007 job, in the ADAMEK application it was in to the Czech language, we have tried to synkopa Syncope (disorder) 271594007

94.8 %. Total cholesterol has been code the attributes of MDMC by means of klaudikace Claudication (finding) 275520000 recorded in 83.4 % of the ADAMEK this classification. In Czech the hmotnost On examination – weight NOS 162770007 medical reports, in free text reports it was abbreviation for this classification is MKN- (finding)

70.4 %. Presence or absence of diabetes 10. Table 3 shows a comparison of MDMC Height and weight (observable entity) 162879003 mellitus was recorded in the software attributes coded by means of ICD-10 výška Body height measure (observable 50373000 application in 95.9 %, in the free text (MKN-10) and SNOMED CT. entity) reports it was 40.7 %. Glycaemia was tělesná teplota Body temperature finding 105723007 recorded in 77.6 % of reports from the Body temperature (observable entity) 276535009 ADAMEK application and in 51.9 % in free obvod pasu Abdominal girth measurement 48094003 text medical reports. Cholesterol was

As the very title of the International Classification of Diseases shows this Tab. 3. Selected attributes of MDMC coded by means of ICD-10 (MKN-10) and SNOMED classification can be used to encode CT. (Part I. - Allergy). particular diseases, syndromes, pathological conditions, injuries, difficulties Attributes of and other reasons for the contact with MDMC Term in MKN-10 English SNOMED CT healthcare services, i.e. the type of information that is being registered by a (in Czech) MKN-10 code equivalent (Concept ID) physician. Unfortunately, using this classification we cannot encode many alergie alergie T78.4 allergy not found attributes of the Minimal Data Model for Cardiology, such as marital status, přítomna manifested education, mental stress, physical stress, alergie na léky alergie na lék T88.7 drug allergy 416098002 physical activity, smoking, alcohol drinking, physical examination (weight, height, body (disorder) temperature, BMI, WHR, etc.), laboratory tests (total cholesterol, HDL-cholesterol) allergic reaction 416093006 or a description of ECG. ICD-10 can be to drug used only for the parts of MDMC related to personal history and current difficulties of a (disorder) possible cardiological origin (see Table 3).

5.7 Standardization of Clinical Contents Tab. 3. Selected attributes of MDMC coded by means of ICD-10 (MKN-10) and SNOMED Attributes, from the point of view of CT. (Part II - Personal history). possibilities of their mapping to standard coding systems, can be classified in the Attributes of following way: — Trouble-free attributes - i.e. attributes, MDMC Term in MKN-10 English SNOMED CT which can be mapped in a direct way, only one possibility of mapping exists, (in Czech) MKN-10 code equivalent (Concept ID) possibly there are only synonyms with diabetes typu E10.- diabetes mellitus 46635009 exactly same meanings and therefore the same classification code (e.g. I type 1 (disorder) patient first name, current smoker, motility, height of a patient, etc.). inzulin E10.- insulin-treated 237599002 — Partially problematic attributes - i.e. attributes, which can be mapped in a diabetes mellitus dependentní non-insulin- way that there are several possibilities dependent of mapping to different synonyms, which differ slightly in their meanings diabetes mellitus and usually in their classification codes (e.g. ischemic cerebro-vascular stroke, (disorder) angina pectoris, hypertension, congestive cardiac failure, etc.). hypertenze Esenciální I10 essential 59621000 — Attributes with a too small granularity, (primární) hypertension i.e. attributes describing certain characteristics on a too general level hypertenze (disorder) so that classification systems contain only terms of a narrower meaning (e.g. ischemická ischemie I25.9 ischemic heart 414545008 e-mail in MDMC versus e-mail to work / e-mail to home / e-mail of a physician choroba srdeční koronární disease (disorder) and so on in classification systems).

Tab. 3. Selected attributes of MDMC coded by means of ICD-10 (MKN-10) and SNOMED CT. benefits to physicians, patients, (Part III - Current difficulties of a possible cardiological origin). administrators, software developers and payers. The standardized clinical Attributes of terminology would help healthcare providers in a way that it could provide MDMC Term in MKN-10 English SNOMED CT complete and easily accessible information that belongs to the process of healthcare (in Czech) MKN-10 code equivalent (Concept ID) (patient's medical record, diseases, treatments, laboratory results, etc.) and it dušnost dušnost R06.8 asthma (disorder) 187687003 would result in better care of patients. bolest na hrudi bolest R07.4 dull chest pain 3368006 Despite problems in the usage of hrudníku (finding) international nomenclatures and matathesauri in healthcare in the Czech palpitace palpitace R00.2 (palpitations) or 161965005 Republic remain, their use is a necessary first step to enable interoperability of (srdce) (awareness of heterogeneous systems of health records. Sufficient semantic interoperability of these heartbeat) or systems is the basis for shared care, which leads to efficiency in healthcare, financial (fluttering of savings and reduction of the burden on heart) patients, and therefore in this work we have tried to analyze how the international synkopa synkopa R55 syncope 271594007 classification systems could be used best for the needs of Czech healthcare. srdeční (disorder) Current healthcare information systems kašel kašel R05 cough 158383001 enable to collect a variety of clinical information, these systems are linked to hemoptýza hemoptýza R04.2 haemoptysis 158384007 clinical knowledge databases, they can search for data, collect data, analyze data, exchange data and they also have a lot of — Attributes with a too big granularity, i.e. standard ones. In special cases it is other functions. As the best classification attributes describing certain possible to add a certain term to an system seems to be SNOMED CT, which characteristics on such a narrow level upcoming new version of a certain coding can provide a basis for these functions. so that classification systems contain system. In case it is not possible to use any Information systems can use the concepts, only a term of a more general meaning of the above-mentioned possibilities of hierarchies and relationships as a common (e.g. symmetrical pulse of carotids, solving mapping problems, it is necessary reference point. SNOMED CT may even etc.). to cope with the fact that mapping will exceed the direct care of patients. This — Attributes, which cannot be found in never be 100%. The insufficient mapping terminology may, for example, facilitate classification systems, e.g. process limits the interoperability of decision support, statistical processing, dyslipidemy, etc. heterogeneous systems used for various monitoring of public health, medical purposes in healthcare. Restricted research and cost analysis. Close cooperation with physicians is interoperability is often inevitable from the essential for solving of such mapping very root of the problem, e.g. insufficient Acknowledgment problems. It is often needed to choose the harmonization of clinical contents of The paper was supported by the projects: right synonym substituting a certain heterogeneous systems of electronic SVV-2010-265513, 1M06014 project of the technical term. It is necessary to do it very health records. Ministry of Education, Youth and Sports CR carefully not to lose information or not to and by the AV0Z10300504 project of the misinterpret it. In case it is not possible to 6. Conclusion Institute of Computer Science AS CR. do it without any lost of information, the The analysis of medical reports written in better way is to describe a non-coded term a free text has shown that recording using by means of a set of several coded terms, a free text is very inhomogeneous and not possibly with showing mutual semantic standardized. The biggest problems for relations. If this is not possible, we can computational processing are typing polemize with specialists whether these errors, various length of shorten “indescribable” terms (attributes) can be expressions and usage of synonyms. The replaced by other more equivalent or more standardized terminology would bring

References Metathesauruses in Shared Healthcare in [22] Adášková J., Anger Z., Asche-rmann M., [1] Bleich H. L., Slack W. V.: Reflections on the Czech Republic. Acta Informatica Bencko V., Berka P., Filipovský J., Goláň electronic medical record: When doctor will Medica, 2005 (13),, 201-205 L., Grus T., Grünfeldová H., Haas T., use them and when they will not, Int. J. Med. [11] Přečková P., Zvárová J., Špidlen J.: Hanuš P., Hanzlíček P., Holcátová I., Inform. 79(2010) 1-4. International Nomenclatures in Shared Hrach K., Jiroušek R., Kejřová E., [2] Hoerbst A., Kohl C. D., Knaup P., Healthcare in the Czech Republic. In: Kocmanová D., Kolář J., Kotásek P., Ammenwerth E.: Attitudes and behaviors Proceedings of 6th Nordic Conference on Králíková E., Krupařová M., Kyloušková related to the introduction of electronic health eHealth and Telemedicie „From Tools to M., Malý M., Mareš R., Matoulek M., records among Austrian and German Services“ (Ed.: Doupi P.), 2006, 45-46 Mazura I., Mrázek V., Novotný L., citizens, Int. J. Med. Inform. 79(2010) 81-89. [12] Elkin P. L., Trusko B. E., Koppel R., Speroff Novotný Z., Pecen L., Peleška J., Prázný [3] Rinner C., Janzek-Hawlat S., Sibinovic S., T., Mohrer D., Sakji S., Gurewitz I., Tuttle M., Pudil P., Rameš J., Rauch J., Duftschmid G.: Semantic Validation of M., Brown S. H.: Secondary Use of Clinical Reissigová J., Rosolová H., Rousková B., Standard-based Electronic Health Record Data. In Seamless Care Safe Care. IOS Říha A., Sedlak P., Slámová A., Somol P., Documents with W3C XML Schema. Press, 2010 (Eds. Blobel B., Hvannberg E., Svačina Š, Svátek V., Šabík D., Šimek S., Method Inf Med 49 (2010), preprint online Gunnarsdóttir), 14-29 Škvor J., Špidlen J., Štochl J., Tomečková [4] Oemig F., Blobel B.: Semantic [13] Stausberg J., Lehmann N., Kaczmarek D., M., Umnerová V., Zvára K., Zvárová J.: A Interoperability Adheres to Proper Models Stein M.: Reability of diagnose coding with proposal of the Minimal Data Model for and Code Systems: A Detailed Examination ICD-10. Int. J. Med. Inform. 2008 (77), 50- Cardiology and the ADAMEK software of Different Approaches for Score Systems. 57 application (in Czech). Internal research Methods Inf Med 2010 49 2, 148-155 [14] IHTSDO: The International Health report of the EuroMISE Centre Cardio. [5] Lopez D.M., Blobel B.: A development Terminology Standards Development Orga- Institute of Computer Science AS CR. framework for semantic interoperable nization: SNOMED Clinical Terms® User Prague, October 2002. health information systems. Int J Med Guide 2008 [23] Mareš R., Tomečková M., Peleška J., Inform 2009; 78 (2), 83-103 [15] Cornet R.: Definitions and Qualifiers in Hanzlíček P., Zvárová J.: Interface of [6] Garde S., Knaup P., Hovenga E.J.S., Heard SNOMED CT. Methods Inf Med 2009 (48), patient database systems - an example of S.: Towards Semantic Interoperability for 177-183 the application designed for data Electronic Health Records: Domain [16] Schulz S., Hanser S., Hahn U., Rodgers collection in the framework of Minimal Knowledge Governance for openEHR J.: The Semantics Procedures and Data Model for Cardiology (in Czech). Cor Archetypes. Methods Inf Med 2007; 46 (3), Diseases in SNOMED® CT. Methods Inf et Vasa, 2002, 44 332-343 Med 2006 (45), 354-8 ( 4), Suppl., 76 [7] Nagy M., Hanzlíček P., Přečková P., Kolesa [17] Gault Lora V., Schultz M.: Variations in [24] Eryiğit G., Nivre J., Oflazer K.: P., Mišúr J., Dioszegi M., Zvárová J.: Medical Subject Headings (MeSH) Dependency Parsing of Turkish. Building Semantically Interoperable EHR mapping: from the natural language of Computational Linguistics. 2008, 34(3), Systems Using International Nomenclatures patron terms to the controlled vocabulary 357-389. and Enterprise Programming Technique. In of mapped lists. J Med Libr Assoc. 2002 [25] Ringlestetter C., Schulz K. U., Mihov S.: eHealth: Combining Health Telematics, April; 90(2): 173180 Orthographic Errors in Web Pages: Toward Telemedicine, Biomedical Engineering and [18] Khan A. N., Griffith S. P., Moore C., Cleaner Web Corpora. Computational Bioinformatics to the Edge. Amsterdam: Russell D., Rosario A. C., Jr., Bertolli J.: Linguistics. 2006, 32(3) 295-340 IOS Press, 2008 (Eds. Blobel, B.; Pharow, Standardizing Laboratory Data by P.; Zvárová, J.; Lopez, D.) 105-110 Mapping to LOINC. J Am Med Inform Contact [8] Nagy M., Hanzlíček P., Přečková P., Říha Assoc. 2006 MayJun; 13(3): 353355 [19] Louis D. N., Ohgaki H., Wiestler O. D., Mgr. Petra Přečková A., Dioszegi M., Seidl L., Zvárová J.: Centre of Biomedical Informatics, Semantic Interoperability in Czech Cavenee W. K., Burger P.C., Jouvet A., Healthcare Environment Supported by HL7 Scheithauer B.W.,Kleihues P.: The 2007 Department of Medical Informatics, version 3. Methods Inf Med 2010 49 (2), WHO Classification of Tumours of the Institute of Computer Science AS CR, 186-195 Central Nervous System. Acta v.v.i. [9] Zvárová J., Hanzlíček P., Nagy M., Neuropathol. 2007 August; 114(2): 97109 Pod Vodárenskou věží 2 Přečková P., Zvára K., Seidl L., Bureš V., [20] Brierley J.: The evolving TNM cancer 182 07 Prague 8 staging system: an essential component Šubrt D., Dostálová T., Seydlová M.: Czech Republic Biomedical Informatics Research for of cancer care. CMAJ. 2006 January 17; Individualized Life-long Shared Healthcare. 174(2): 155156 e-mail: [email protected] In: Biocybernetics and Biomedical [21] Campbell J.R., Olivek D.E., Shortliffe: Engineering, 2009, 29 (2), 31-41 UMLS: towards a collaborative approach [10] Přečková P., Špidlen J., Zvárová J.: Usage for solving terminological problems, J. of the International Nomenclatures and Am. Med. Inform. Assoc. 5 (1998), 12-16