Dimitar Tcharaktchiev

Text Mining and Big Data Analytics for Retrospective Analysis of Clinical Texts from Outpatient Care

Abstract This paper presents the results of an on-going research project for knowledge extraction from large corpora of clinical narratives in Bulgarian language, approximately 100 million of outpatient care notes. Entities with numerical values are mined in the free text and the extracted information is stored in a structured format. The Algorithms for retrospective analyses and big data analytics are applied for studying the treatment and evaluating the diabetes compensation and control of arterial blood pressure.

Applying Language Technologies on Healthcare Patient Records for Better Treatment of Bulgarian Diabetic Patients

This paper presents a research project integrating language technologies and a business intelligence tool that help to discover new knowledge in a very large repository of patient records in Bulgarian language. The ultimate project objective is to accelerate the construction of the Register of diabetic patients in Bulgaria. All the information needed for the Register is available in the outpatient records, collected by the Bulgarian National Health Insurance Fund. We extract automatically from the records’ free text essential entities related to the drug treatment such as drug names, dosages, modes of admission, frequency and treatment duration with precision 95.2%; we classify the records according to the hypothesis “having diabetes” with precision 91.5% and deliver these findings to decision makers in order to improve the public health policy and the management of Bulgarian healthcare system. The experiments are run on the records of about 436,000 diabetic patients.

Mining comorbidity patterns using retrospective analysis of big collection of outpatient records

BackgroundStudying comorbidities of disorders is important for detection and prevention. For discovering frequent patterns of diseases we can use retrospective analysis of population data, by filtering events with common properties and similar significance. Most frequent pattern mining methods do not consider contextual information about extracted patterns. Further data mining developments might enable more efficient applications in specific tasks like comorbidities identification.MethodsWe propose a cascade data mining approach for frequent pattern mining enriched with context information, including a new algorithm MIxCO for maximal frequent patterns mining. Text mining tools extract entities from free text and deliver additional context attributes beyond the structured information about the patients.ResultsThe proposed approach was tested using pseudonymised reimbursement requests (outpatient records) submitted to the Bulgarian National Health Insurance Fund in 2010–2016 for more than 5 million citizens yearly. Experiments were run on 3 data collections. Some known comorbidities of Schizophrenia, Hyperprolactinemia and Diabetes Mellitus Type 2 are confirmed; novel hypotheses about stable comorbidities are generated. The evaluation shows that MIxCO is efficient for big dense datasets.ConclusionExplicating maximal frequent itemsets enables to build hypotheses concerning the relationships between the exogeneous and endogeneous factors triggering the formation of these sets. MixCO will help to identify risk groups of patients with a predisposition to develop socially-significant disorders like diabetes. This will turn static archives like the Diabetes Register in Bulgaria to a powerful alerting and predictive framework.

Mining Clinical Events to Reveal Patterns and Sequences

This paper presents results of ongoing project for discovering complex temporal relations between disorders and their treatment. We propose a cascade data mining approach for frequent pattern and sequence mining. The main difference from the classical methods is that instead of applying separately each method we reuse and extend the result prefix tree from the previous step thus reducing the search space for the next task. Also we apply separately search for diagnosis and treatment and combine the results in more complex relations. Another constraint is that items in sequences are distinct and we have also parallel episodes and different time constraints. All experiments are provided on structured data extracted by text mining techniques from approx. 8 million outpatient records in Bulgarian language. Experiments are applied for 3 collections of data for windows with size 1 and 3 months, and without limitations. We describe in more details findings for Schizophrenia, Diabetes Mellitus Type 2 and Hyperprolactinemia association.

Medical Archetypes and Information Extraction Templates in Automatic Processing of Clinical Narratives

This paper discusses the notion of medical archetype and the manner how the archetype elements are documented in hospital patient records. This is done by interpreting the archetypes as information extraction templates in automatic text analysis of clinical narratives. The extensive extraction experiments performed over thousands of anonymous discharge letters show the actual instantiation of the required and expected items in the narrative clinical documentation; in fact much tacit medical knowledge is implicitly presented in the real clinical texts. This fact suggests that the archetype approach to defaults and inheritance might need certain development.

Ethics and security in text mining of patient records in Bulgarian: the EVTIMA solution

This paper presents the current solutions concerning ethical and security issues in the system EVTIMA -- an environment supporting text mining of patient records (PRs) in Bulgarian. Confidentiality and anonymisation of the analysed documents are crucial from ethical point of view and are considered as leading development principles. Here we describe in detail our method for PR de-identification of PRs which uses data vocabularies, regular expressions and additional heuristics to locate the identification information. It is trained on a corpus of 197 documents and tested on 1000 documents. The algorithm works in three steps and de-identifies 97% of the personalising information. Thus it is comparable to the reported results in similar tasks in English.

EVTIMA: a system for IE from hospital patient records in Bulgarian

In this article we present a text analysis system designed to extract key information from clinical text in Bulgarian language. Using shallow analysis within an Information Extraction (IE) approach, the system builds structured descriptions of patient status, disease duration, complications and treatments. We discuss some particularities of the medical language of Bulgarian patient records, the architecture and functionality of our current prototype, and evaluation results regarding the IE tasks we tackle at present. The paper also sketches the original aspects of our IE solutions.

Medical, social, and economic consequences of type 2 diabetes therapy with medicinal products from the group of DPP-4i, SGLT-2i, and GLP-1 RA

Incretins (DPP-4i and GLP-1 RA) as well as SGLT-2i comprise a broad range of pharmacological groups for type 2 diabetes treatment with not well established and documented long-term therapeutic and economic effect. We aimed to analyze the changes in HbA1c level and related risk of diabetes incidents appearance, as well as to value the cost and results of incretins and SGLT-2i-based therapy in Bulgaria. Information about the changes in the HbA1c level was extracted for 705,515 type 2 diabetic patients from the National diabetes register during 2012–2016. Patients treated with DPP-4i (sitagliptin, vildagliptin, saxagliptin, and linagliptin); SGLT-2i (dapagliflozin and empagliflozin); GLP-1 RA (exenatide, liraglutide, lixisenatide, andexenatide extended-release); DPP-4i + МЕТ (sitagliptin/metformin; vildagliptin/metformin, saxagliptin/metformin, andlinagliptin/metformin); and SGLT2i + МЕТ (dapagl i f loz in /met formin and empagliflozin/metformin) were selected. For 10,547 patients, information was found about the initial and final HbA1c values, and 6122 performing a decrease in the HbA1c level were analyzed. Literature evidences about the decrease in the relative risk (RR) and number of diabetic incidents in case of 1% decreases in HbA1c were collected [1, 2]. After the therapy, 3356 people are having less than 7% HbA1c that is considered as very good diabetic control. The number of people with HbA1c above 8% is decreasing significantly (n = 1453) and above 9% is decreasing almost four times. HbA1c reduces its level with highest percentage for patients treated with GLP-1 RA (1.76%), followed by those treated with DPP-4i (1.43 for monotherapy and 1.71 for combination with metformin) and SGLT-2i. The reduction for diabetic incidents in case of 1% reduction in HbA1c level is between 20 for fatal and nonfatal stroke to 62 for microvascular complications. Total average reduction is highest for patients on GLP-1 RA (n = 347 incidents for patients on monotherapy and 337 incidents for patients on combination therapy), and those on DPP4i +MET (n = 337

Data Mining and Analytics for Exploring Bulgarian Diabetic Register

This paper discusses the need of building diabetic registers in order to monitor the disease development and assess the prevention and treatment plans. The automatic generation of a nation-wide Diabetes Register in Bulgaria is presented, using outpatient records submitted to the National Health Insurance Fund in 2010–2014 and updated with data from outpatient records for 2015–2016. The construction relies on advanced automatic analysis of free clinical texts and business analytics technologies for storing, maintaining, searching, querying and analyzing data. Original frequent pattern mining algorithms enable to discover maximal frequent itemsets of simultaneous diseases for diabetic patients. We show how comorbidities, identified for patients in the prediabetes period, can help to define alerts about specific risk factors for Diabetes Mellitus type 2, and thus might contribute to prevention. We also claim that the synergy of modern analytics and data mining tools transforms a static archive of clinical patient records to a sophisticated knowledge discovery and prediction environment.

From Individual EHR Maintenance to Generalised Findings: Experiments for Application of NLP to Patient-Related Texts

Experiments in automatic analysis of free texts in Bulgarian hospital discharge letters are presented. Natural Language Processing (NLP) has been applied to medical texts since decades but high-quality results have been demonstrated only recently. The progress in automatic text analysis opens new directions for secondary use of Electronic Health Records (EHR). It enables also the design and development of software systems which provide better patient access to his/her health records as well as better maintenance of large EHR archives. We report about successful extraction of important patient-related entities from hospital EHR texts and consider several scenarios for application of NLP modules in healthcare software systems.