REPORT A systemic biologic model for healthcare data quality

REPORT A systemic biologic model for healthcare data quality Hamid Moghaddasi and Forough Rahimi Poor documentation of healthcare data, including the recording of inaccurate and false information, is a concern for health data collections at all levels, whether large hospitals or small clinics (WHO 2003). The National Committee on Vital and Health Statistics (Meyer 2000; Center for Health Information Quality 2002) stated that high quality healthcare depends on access to accurate and comprehensive medical records. Such information is essential for effective diagnosis and treatment, measurement and improvement of healthcare quality, advancement of public health, improvement of healthcare productivity, and facilitation of cost reimbursement. Rigby et al. (1998) commented on the beneicial effects of high quality information on healthcare quality, arguing that data quality is at the very ‘heart’ of medicine. Providing high quality healthcare, using available resources eficiently, and providing ongoing high quality services that satisfy the needs of society can be achieved only through effective communication and public responsibility to resolve patients’ problems. Without reliable information, high quality healthcare is dificult to achieve (Shortliffe & Cimino 2014). Correct, timely and accessible healthcare data play a signiicant role in developmental planning and support of healthcare services. It is essential to maximise quality and timely distribution of data to facilitate provision of health care to individuals at an optimal level (irrespective of the level at which services are provided). Data quality is also important in monitoring performance of healthcare institutions and staff. Data collected and utilised should be accurate, complete, reliable, and accessible to authorised users, including doctors, healthcare teams, healthcare institutions, legal authorities, and state, province, and national governmental healthcare authorities. WHO (2003) announced the importance of data quality, with accurate and reliable healthcare data required for the following cases: ƒ continuing future patient health care at all levels of healthcare ƒ medico-legal purposes for patients, physicians and healthcare institutions ƒ providing correct accurate and reliable information about diseases treated and surgical procedures performed in a hospital and in the society together with immunisation and screening programs, including the number and type of participants ƒ clinical and healthcare services research and healthcare intervention outcomes ƒ complete, correct, accurate and reliable statistical data on the use of healthcare services in the society ƒ training healthcare professionals 28 ƒ designing employment requirements and planning healthcare services. Literature The healthcare environment is dependent on computerised information systems and the amount of collected, used and stored data is increasing. Computerised systems and related databases are faster and larger than a decade ago. Systems are more operational and complex, providing access to data directly and immediately. However, larger, faster and more complex systems are not necessarily better. Remember the classic example about computer systems: if it’s garbage in, then it’s garbage out. Entering low quality data into a system that is getting larger and more complex will only generate more garbage, which will then be distributed to a wide range of users. Data with inherent quality are those that are comprehensive, current, relevant, accurate, timely, and appropriate. This raises the question: ‘What are the characteristics of data quality in a healthcare environment?’ The Institute of Medicine (IOM) in the US has reported results of numerous studies investigating different aspects of data quality that indicate poor documentation of patients’ medical records. The importance of the quality of collected, stored, and processed data intensiies when considerable amounts of erroneous data are entered into information systems that are rapidly becoming faster, larger and more complex. Fundamental questions are: ƒ Can we protect our healthcare databases from the impact of low-quality data? ƒ Can organisations estimate their costs based on incomplete, inconsistent, or unreliable data? ƒ Can care providers make appropriate clinical decisions if patients’ records are incomplete, unhelpful, and unreliable or lack necessary data? ƒ Can public health institutions carry out their projects, identify and prevent diseases without current comprehensive, accurate, and relevant data? ƒ Can we monitor the quality and clinical care outcomes without data repositories that are necessary, current, appropriate, accurate, complete, reliable, comprehensive, and relevant? High quality data lead to commercial or strategic success in the marketplace. It is essential that healthcare organisations strive to maintain high quality data and develop processes, policies, and guidelines to protect the value of the data (Johns 2002; Davis & Lacour 2002). According to Johns, characteristics of data quality are often explained in terms of relevancy, completeness, accuracy, precision and accessibility. More often, HIM-INTERCHANGE Vol 6 No 1 2016 ISSN 1838-8620 (PRINT) ISSN 1838-8639 (ONLINE) ‘integrity’ is used as a generic term to cover all characteristics of data quality. One of the most complete views about aspects of data quality was presented by Redman (1992, cited in Johns 2002). Redman’s model was based on assessing characteristics of the end users’ sub-schema or conceptual views of the data. In other words, data quality (as with other types of quality), is an entity that is deined by incorporating the conceptual view of end users of the data. Discussion Characteristics or attributes of data quality used in this article correspond with views of experts and accredited organisations that analyse their meaning and concept. Accuracy speciied timeframe. So, in this sense one of the meanings of ‘timeliness’ is ‘currency’. Completeness Completeness can be interpreted as ‘the existence of all the necessary data’ (WHO 2003), and each mandatory data element within a data set should be completed, even if data entry is delayed because of (for example) unforeseen emergency circumstances. The data element is a guide term that is explained by a clear, obvious, and standard words and it is data absorbent. Words such as ‘name’, ‘sex’, ‘age’, ‘description of operation’, ‘cause of death’, and ‘main diagnosis’ are all data elements, each of which absorbs the related data (Moghaddasi 2009). Johns (2002) and Abdelhak (2010) explained the necessity of recording all essential data through the expression of ‘comprehensiveness’. They did not consider ‘completeness’ as the characteristic of data quality; rather, it was a prerequisite for data quality. Johns suggested that necessary data should be collected based on a speciic domain or scope that is referred to as ‘data set’. Data should be accurate. If a patient’s gender is male, it must be documented as male in his record. If the patient’s name is James Russell, the same name must be written in full in the record. Data in any format must accurately relect what really happened (Johns 2002). Clearly (2001, cited in Moghaddasi et al. 2014) recommended involving patients in conirming the accuracy of some, especially demographic, data. Relevancy According to Ippilito (cited in Orli 1996), accuracy is the Johns (2002) emphasised the importance of ‘data relevancy’. degree of agreement between the value of a datum and a This characteristic indicates that the meaning of the data source that is assumed to be correct. Ippilito deined it as ‘a should predicate the implementation of the process or the qualitative assessment of freedom from error’. Johns (2002) application for which the data are collected. One of the rejected Ippilito’s view, arguing that accuracy and the deterprocesses covered at admission is to collect demographic mination of the degree of accuracy are two distinct concepts. data of patients so that they can be distinguished from each To determine the accuracy of a data element its value needs other. Recording a patient’s name, date of birth, and gender to be veriied. How can this be achieved in the absence of the is appropriate, while data such as their leisure activities and data source? We can only hope names of their pets are irrelevant that accuracy of data will increase . Once the conceptual view of the If there is not a signiicant relationship through implementing database user has been developed, all data between the data and the process or management techniques, such as should be distinguished as either application for which the data are collected, relevant or unnecessary. Abdelhak checking for consistency in the data and limitations of the values the data will not be useful, usable, or it for (2010) emphasised the need for a of the ield. signiicant relationship between data purpose and the process or application for which these data are collected. The Timeliness International Standards Organization (ISO) (2008) identiied The need for timely data depends on context, that is, the this as being ‘it for purpose’. If there is not a signiicant relasituation and the conceptual view of the user (Johns 2002). tionship between the data and the process or application for In the ICU of a hospital, timeliness of data is so important which the data are collected, the data will not be useful, usable, that we may need to collect and interpret data every second, or it for purpose. whereas data used in a patient’s regular physical examination may not be as time-sensitive. However, timely registration of these data still plays an important role in effective treatment Consistency of patients (Clearly 2001, cited in Moghaddasi et al. 2014), When different entities have common or similar attributes, and entering test results into the computer, registration of it is expected that the value of the attributes will also be diagnosis and reporting surgery on time generates usable identical. The number of a patient’s medical record must information. Clearly also recommended that deadlines for data be the same in all reports during a care episode. A type of entry be based on national consensus and not be subjective, inconsistency occurs when two related, but not identical, while also acknowledging that recording of data should not information entities do not match. For example, ‘hysterectomy’ delay urgent treatment of a patient. WHO (2003) emphasised and ‘gender’ are related terms and this surgery is relevant only that information, especially clinical information, and treatment to women. There is inconsistency if it is recorded that hysteror results of treatment, should be documented immediately ectomy is performed on a male patient. as delays result in omissions and errors. Ippilito (cited in Ippilito (cited in Orli 1996) maintained that consistency Orli 1996) deined this characteristic of data quality as the exists when data do not contradict each other. Consistency extent to which data elements can be made available within a can be viewed as internal correspondence between individual HIM-INTERCHANGE Vol 6 No 1 2016 ISSN 1838-8620 (PRINT) ISSN 1838-8639 (ONLINE) 29 REPORT data and data elements should be deined so that both current and future users could understand them datum elements and the data. As consistency emphasises internal consistency of an information entity, the existence of consistency and a logical relationship between elements of data, it follows that inconsistency must be due to inaccuracy. It can be argued that there is no difference between the two characteristics of consistency and accuracy, and the presence or absence of one means the presence or absence of the other. The Center for Health Information Quality (2002) also identiied accuracy, relevancy and clearness as the three main components of data quality, with consistency considered a component of accuracy. Deinition Abdelhak (2010) believed that data and data elements should be deined so that both current and future users could understand them. Each data element should have a clear meaning and an acceptable value (AHIMA 2009). A concise and clear deinition of data and data elements facilitates accurate data collection. Completely clear and illustrative deinitions for the data and the determination of acceptable values for them result in universal understanding of the data and data elements. Some experts consider the characteristics of ‘uniqueness’ and ‘precision’ to be independent attributes of data quality, but according to the meaning of these terms, they are actually components of its deinition. Ippilito (cited in Orli 1996) considered uniqueness as a key value of data. Johns (2002) also maintained that acceptable values or range of values should be determined for each attribute and values should be suficient to support relevant application or process. This characteristic of data quality is precision. For example, regarding the thickness of the needle used in a catheter, the precision or scope of values range between 16 and 22; and for the amount of insulin prescribed, the precision or the scope of values range from 1 to 12. Providing data in as much detail as possible helps to reduce the risk of error. Abdelhak (2010) conirmed John’s view that acceptable values be determined for each data element. For instance, the determined values for gender are male, female, and unknown. Both Johns (2002) and Abdelhak (2010) considered the more accurate and detailed explanation of data and their values as a characteristic of data quality called ‘granularity’, and that these characteristics and data values should be determined correctly and precisely. A patient’s body temperature should be recorded to the precision of a tenth of a degree; there is a signiicant physiological difference between, say, a temperature of 101.1°C and 101.9ºC. Hence, values of data such as the patient’s body temperature should be determined precisely 30 and accurately. As a result, the concept of ‘granularity’ emphasising the more accurate and precise data and their values is considered as a component of ‘deinition’. Reliability This feature emphasises the need to repeat the data collection, processing, storage, and the representation of the data; consistent results depend on the consistency of the input data. As discussed earlier, the concept of ‘consistency’ implies ‘accuracy’; and the concept of ‘accuracy’ covers ‘reliability’. Thus, according to the deinition provided by the organisation, this feature depends on ‘consistency’ or ‘accuracy’. It should also be mentioned that according to the above deinition, this characteristic is a combination of the following attributes: ‘deinition’, ‘completeness’, ‘accuracy’, ‘relevancy’ and ‘timeliness’. Coverage The concept of ‘coverage’ was proposed by Clearly (2002) and according to this characteristic, data should relect whatever is done by the healthcare institution. Regardless of Clearly’s speciic deinition based on what he maintained was the attribute of ‘completeness’, it can be understood that the characteristic of ‘coverage’ implies the concept of ‘completeness’. Comparability The feature of ‘comparability’, also introduced by Canadian Institute for Health Information (Canadian Institute for Health Information 2009; AHIMA 2003) is a function of the characteristic of ‘deinition’ of the data. If a datum or a fact is not meaningful, it cannot be used to compare or justify similarities or differences in entities. The characteristic of ‘deinition’ creates the attribute of ‘precision’ of the data and causes the data to be meaningful and understandable. Meaningfulness of data causes the characteristic of ‘comparability’ of data to appear. Validity Davis (2002) believed that data recorded in any format may be affected by human error. To be useful, data must be valid. The characteristic of ‘validity’ implies that data are in accordance with acceptable and expected scope. In the American postal system «ABCDE» is not a recognised postal code because US postal codes only include digits. AHIMA (2009) maintained that data should be in an agreed format and based on national standards. Clearly (2002) considered a need for data to be in accordance with HIM-INTERCHANGE Vol 6 No 1 2016 ISSN 1838-8620 (PRINT) ISSN 1838-8639 (ONLINE) national standards as the ‘validity of data’. As the characteristic of ‘deinition’ represents clear and illustrative explanations, determining an acceptable range of values for data, and data elements (in an agreed format and based on national or international standards), it seems that the characteristic of the ‘deinition’ covers ‘validity’. The ‘security of data’ emphasises secure storage in order not to harm the data. ‘Conidentiality’, which is the patient’s right, represents the internal and external disclosure policies of the organisation. It implies conditional release of information. Conclusion Data representation format Redman (1992, cited in Johns 2002) added this feature to the set of characteristics based on the conceptual view of the end user to evaluate data quality. It is a set of rules used to record data. The value of the datum should be represented in a speciic format and through a speciic language. As the main core of data quality, the data representation format includes several characteristics such as type of language (character, symbol, picture), the size of character, symbol or picture, font style, ink, color, rules, margin, horizontal or vertical spacing. The most important characteristic is legibility; all data, whether handwritten, typed, or printed must be legible. According to Johns (2002), ‘representation format’, the format through which data are represented to the end user, affects data quality. By format we do not only mean the size of a computer screen or printed reports; we also mean the language (symbols) used to transfer the meaning of the data. Laboratory results usually include characters. Results may also be highlighted (when data are not in the normal range). An example of the graphic symbols used in the medical documentation is ♂ (male) and ♀ (female). Using metaphors in the representation of data value can be useful. The representation format functions as ‘body’ for the nature or content of data, without which data cannot be tangible, as the soul needs a body through which to communicate. Unlike Johns (2002), who argued that representation format is different from characteristics of data, it can be argued that without the representation format data are meaningless and even if we consider these data meaningful (which is impossible), they cannot represent meaning without the representation format. Humans use their minds to sort and process data as mediators or observers, and people often need to transfer mental images to words, gestures, and even visualisation in order to communicate. Thus, data are a series of content and representational formats. Content is like the soul and the representation format is like its body. In studies covered in this article, 24 attributes or characteristics from various sources originating in the US, Canada, the UK, Australia and WHO were proposed for the quality of healthcare data. Some attributes were found to result from one or more main features, while others were the result of combining a number of key features. Some were not related to factors forming the nature of the data; they were associated with information management processes and were considered as the quality of information management. Redman (1992) introduced a set of features including accuracy, timeliness, currency, precision, relevancy, consistency, comprehensiveness, and granularity based on the conceptual view of the end user, and added data representation format to the list. Thus it became known as Redman’s model. The Center of Health Information Quality (2002) also introduced a model consisting of three core elements, including: accuracy, relevancy, and clearness, each of which has its own component(s): ƒ accurate (consistent, continuity, current, reliable) Ambiguity in concepts It is worth mentioning that accessibility, security and conidentiality, introduced by some organisations and individuals as characteristics of data quality, are not deinitively related to factors forming the nature of data and cannot be considered as characteristics of data quality itself. These characteristics are related to the process of storage, retrieval, and distribution of information and are characteristics of the quality of information management. In discussing ‘accessibility’, Abdelhak (2010) maintained that all data should be easily accessible and usable (for all clinical, administrative, and organisational purposes) and collecting these data should be legal. If the data are not available, the value of collecting and documenting them correctly disappears. Figure 1: A systemic biologic model of data quality HIM-INTERCHANGE Vol 6 No 1 2016 ISSN 1838-8620 (PRINT) ISSN 1838-8639 (ONLINE) 31 REPORT ƒ ƒ relevant (accessible, appropriate, patient involvement) clear (appearance of text, presentation, content). In the present study the proposed model, analogous to a biological entity (such as humans), is that data are entities that consist of two main components: content (nature-soul), and representation format, (corpus-body) (Figure 1 refers). The characteristics of data quality are determined based on factors forming each of these two components. The foundation of this model is based on studies conducted by various organisations and individuals. In addition to the ive characteristics of accuracy, completeness, timeliness, relevance, and deinition, which are the result of the analysis of the 24 characteristics in the present study and are related to the content of the data, it is also necessary to add the characteristic of ‘logical linkage’ to each of the two main components of data quality (content and format). By ‘logical linkage’ we mean that there is a strong and logical relationship between the two main components of data quality (content and data representation format), a strong and logical relationship among each of the relevant characteristics of the two components (if needed), and a strong and logical relationship between the characteristics forming each of them. Since the system is a set of components connected and coordinated to achieve a speciic purpose or purposes, the characteristic of ‘logical linkage’ in the set of characteristics of data quality completes the existence of data as a system and improves the data quality as a system. In terms of completeness of data, there should be a relationship and a logical sequence among the necessary data. If we look at a set of the characteristics of deinition, accuracy, completeness, relevancy, and timeliness generally, the strong and logical relationship and coordination among them must be clearly elicited. The absence of this feature shows that the components of the data are torn and the characteristics of data quality are detached from each other. Meyer, C. (2000).Uniform data standards for patient medical records information. Available at: http://www.ncvhs.hhs.gov/wp-content/ uploads/2014/08/hipaa000706.pdf (accessed 27 Nov 2015). Moghaddasi, H., Rabiei, R. and Sadeghi, S. (2014). Improving the quality of clinical coding: a comprehensive audit model. Journal of Health Management and Informatics 1(2): http://jhmi.sums.ac.ir/index.php/ JHMI/article/view/16. Moghaddasi, H. (2009). Health data processing. Tehran,Vajehpardaz. Orli, J.R. (1996). Data quality methods. Available at: http://kismeta.com/ cleand1.html (accessd 27 Nov 2015). Rigby, M., Roberts, R., Williams, J., Clark, J., Savill, A., Lervy, B. and Mooney, G. (1998). Integrated record keeping as an essential aspect of a primary care led health service. British Medical Journal 317(7158): 579–582. Shortliffe, H.E. and Cimino, J.J. (2014). Biomedical informatics. USA, SpringerVerlag. World Health Organization (2003). Improving data quality: a guide for developing countries. Available at: http://www.phinnetwork.org/ portals/0/improving_data_quality.pdf (accessed 27 Nov 2015). References Abdelhak, M. (2010). Health information: management of a strategic resource. USA, W.B. Saunders Company. American Health Information Management Association (AHIMA) (2009). Practice brief: data quality management model. Available at: http:// www.umass.edu/eei/2009Workshop/pdfs/Data%20Quality%20 Management%20Model.pdf (accessed 27 Nov 2015). American Health Information Management Association (AHIMA) (2003). Management and improving data quality. Journal of AHIMA 74(7): 64 A-C. Canadian Institute for Health Information (CIHI). (2009). The CIHI Data Quality Framework, (Ottawa, Ontario.: CIHI. Center for Health Information Quality (CHIQ). (2002).Guidelines for health information quality. Available on: www.hfht.org/chiq. Clearly, S. (2002). The NISHA Hospitals NHS Trust; Data Quality Policy. (Available from the authors upon request). Davis, N. & Lacour M. (2002). Introduction to health information technology. USA, W.B. Saunders Company. International Standards Organization (ISO) (2008). ISO 9001:2008. Quality management systems requirements. Available at: http://www.iso.org/iso/ catalogue_detail?csnumber=46486 (accessed 27 Nov 2015). Johns, L.M. (2002). Information management for health professions. USA, Delmar Publishers. 32 Corresponding author: Hamid Moghaddasi, PhD Associate Professor of Health information Management and Medical Informatics Head of Health information Management and Medical Informatics Department College of Paramedical Sciences Post Code 19395 Shahid Beheshti University of Medical Sciences Darband Street, Ghods Square (Tajrish),Tehran, Iran Moghaddasi@sbmu.ac.ir Forough Rahimi, PhD Assistant Professor, College of Paramedical Sciences Shahid Beheshti University of Medical Sciences Tehran, Iran frahimi@sbmu.ac.ir HIM-INTERCHANGE Vol 6 No 1 2016 ISSN 1838-8620 (PRINT) ISSN 1838-8639 (ONLINE)