Medical Informatics in Ensuring Quality Control of Cancer Care: Promising Directions of Development

Full Text

Introduction
Malignant tumors place a heavy socio-economic burden on society and the clinical health care infrastructure. Determination of effective methods of internal control of the quality of oncological care, carried out, among other things, with the aim of increasing the wide availability of innovative and effective treatment algorithms, is the most important task in the light of the implementation of the Federal project to combat cancer, which provides for a significant reduction in the mortality rate of the population of the Russian Federation from malignant neoplasms by 2024 year.
The rapid evolution of oncoinformatics determines the global digitalization of technologies for managing the organization of internal quality control of oncological practice [1]. Routine clinical data are increasingly analyzed using the approaches of a new branch of knowledge - oncoinformatics, which emerged at the junction of basic sections of informatics, such as: informatics of clinical trials, informatics of imaging research methods, informatics of pathomorphological diagnostics, etc. The development of clinical oncoinformatics is an example of an ideal way to improve tools for creating an effective quality control model based on the principles of evidence-based medicine [2] [3].
Modern methods of processing large data arrays and CancerLinQ technology (fast-learning information and analytical oncological network for quality control and assurance), elements of information systems for mobile and "smart health" [4], can be considered as "breakthrough" technological platforms embedded in the foundation of sustainable development of integrated automated systems for comparative assessments and continuous monitoring of quality in real time [5].
The study of the dominant trends in the development of technologies for clinical oncoinformatics is extremely important and relevant. Thus, the purpose of this review is to summarize innovative directions in oncoinformatics that have the most promising potential for further improving quality control mechanisms in oncology.

Materials and methods
The PubMed (Medline) database was used to search for relevant information. And also the search engine GOOGLE. The research used thematic and semantic technologies of information extraction. In the search box, queries were entered on the topic of quality assurance and clinical oncoinformatics: “oncology informatics”, “cancer care”, “clinical audit”, “auditing”, “big data”, “Dutch Institute of clinical auditing”, “CancerLinQ”, “ ASCO "," CAPTIVE ", IBM etc. Sources were selected "manually" by meaning, automated filters were not applied.

Research results

Prospects for the use of large data arrays in assessing the quality of cancer care

First of all, it should be noted that in the scientific literature under big data in healthcare by the company - the developer of super-high-speed high-flow computers - International Business Machines (IBM) means data that meets the following 5 criteria (rule 5 - V) [6]:
1. Volume: the full range of data on many observations of various patients, including key characteristics of diagnosis, treatment and outcomes, socioeconomic and other parameters;
2. Velocity: a) big data in oncology should be generated with increasing speed; b) calculations and data processing should be done relatively quickly. It is important to ensure that data collection and processing is equally fast in all departments of the oncology network.
3. Variety: large arrays should contain and reflect the heterogeneity and huge variety of data types (classes and clusters) that figure in everyday cancer practice;
4. Variability (variability) of data due to the heterogeneity of methods of collecting information that differ over time, as well as depending on different circumstances and conditions of collection within and between medical organizations (from the temporal and spatial context). In doing so, an accurate interpretation of the data becomes possible in relation to the context.
5. Value - it makes sense to create an infrastructure for the collection and interpretation of big data only if the results of the analysis of this data and conclusions will lead to improvement or have an impact on the organization of health care [6].
The aforementioned variety of properties of big data arrays and the creation of a unified information base of information demonstrating the relationship between the characteristics of patients and the treatment performed with the achieved outcomes opens up unprecedented opportunities for the formation of automated systems of "emergency feedback" with informing medical organizations about the results of multifactorial (including comparative) assessments of "quality and the effectiveness of current cancer practice.
Research shows that a solid foundation has been created to build high-speed analytical computing systems that provide immediate, real-time feedback on reported medical outcomes in oncology. Their goal is to determine the optimal correlations of clinical tests and algorithms performed with outcomes. Interestingly, such assessment tools also stimulate medical staff, including laboratories, to revise and improve the quality of working protocols for the provision of cancer care [2]. Although improved quality control is only possible if transparency is maintained, it should be noted that feedback scores, especially sensitive mirroring data on outcomes and standardized parameters, should be provided with extreme caution, as laboratories and hospitals may be concerned about own reputation [7].
Publication of anonymized / anonymized data in the community and full disclosure of information from the feedback system to the healthcare provider exclusively at the individual level, without wide publication, motivates medical institutions to cooperate and integrate into the system of "mirror / comparative assessments" of clinical activities [2]. An example of a successful organization of this approach is the development of automated feedback in medical electronic resources in the Netherlands. Algorithms for clinical oncoinformatics are implemented with the participation of the Netherlands Institute of Clinical Auditing and are used to form the so-called “quality registers” (Figure 1) [8].

It was agreed that the level of transparency, as a rule, should be increased stepwise: in the first year, medical organizations are only involved in the audit, in the second year, transparency extends to the criteria for evaluating medical processes, then it is possible to increase the level of transparency before posting the results of evaluating treatment outcomes in open resources. Thus, before the results of assessments according to the developed criteria are published in external sources, hospital staff have the opportunity to improve the quality of oncological care in the field by receiving and studying feedback. The last word in the decision to publish data always rests with hospitals; their quality assessment results are not made available to external parties in open sources if the hospital administration has not approved / agreed an appropriate level of transparency [8].
Representatives of scientific associations participating in audits agree once a year which quality assessment criteria are transparent and suitable for publication. They discuss this on so-called "indicator days" with the staff of the Netherlands Institute for Health and Social Care, as well as with representatives of the medical professional community, with representatives of patient organizations, insurance companies and hospitals [8].
Mirror, comparative picture of qualitative and quantitative automated assessments of oncological practice, indicating a higher recurrence rate of malignant tumors in a particular medical organization in relation to advanced expert oncological centers allows a targeted audit of the functioning of the involved health care links and successfully eliminate the causes of deficiencies [2].

CancerLinQ technology in quality control of oncology practice
Under the patronage of ASCO, on the SAP platform, a fast-learning automated electronic network for monitoring and quality assurance of cancer practice CancerLinQ (Cancer Learning Intelligence Network) has been developed [9]. The initiative is supported by medical practitioners and reflects ASCO's core mission of providing quality cancer care to patients [10]. The system takes advantage of big data bases for "learning", analyzing each individual clinical case. The network compares indicators of the processes and outcomes of cancer care against approved treatment standards and generates a quick response to specialists in current clinical practice about the achieved quality values. CancerLinQ's goals include [1]: 1) providing real-time quality assessment feedback when cancer organizations receive rapid assessments of adherence to clinical guidelines and practices of excellence centers, which directly contribute to the improvement of internal quality self-control processes; 2) provision of patient-oriented / individualized information and treatment algorithms for each patient in accordance with clinical guidelines and information in other databases; 3) calculating patterns to improve existing practice, synthesize new scientific hypotheses and improve clinical guidelines. In the future, it is planned to expand the purposes of CancerLinQ in relation to the analysis of real-world data - from the selection of candidates for inclusion in clinical trials, monitoring treatment outcomes, to the study of the efficacy and safety of drugs in everyday clinical practice [1].
The digital oncology network collects data directly from electronic medical records (EHR) and electronic medical management components of attached oncological institutions. The advantage of the network is the automated collection of all possible data from primary sources without prior selection and their transfer to a series of storages according to the structural architecture of CancerLinQ. Briefly: data from electronic databases of medical information or EHR, medical process management databases and other sources enter the "secure" lake storage, which provides a high degree of protection of medical information and personalized data, CancerLinQ transforms (standardization, normalization, ontologization and conceptualization) and moves encrypted data from the primary database ("lake storage") to the secondary storage of processed data [1] [11].
One of the new results of the scientific development of the CancerLinQ system was published by Potter et al. in 2020 [9]. It is expected that with the improvement of the CancerLinQ system that studies aspects of healthcare and with the expansion of the service area (volume and flow of information), the practical applicability of the technology will significantly increase. All this will lead to an increase in the role of real clinical practice data (“real world data”) in the preparation of clinical guidelines, study of drug efficacy and safety, and determination of effective quality control mechanisms based on the principles of evidence-based medicine.

CAPTIVE infrastructure and quality control

More recently, in 2020, Stanford University researchers [12] published a paper demonstrating the applicability of digital data for the automated determination of the quality of oncological practice based on the analysis of primary sources, including electronic medical record (EHR) records [12]. An experimental infrastructure CAPTIVE was developed, technologies for processing natural language variables and machine learning techniques were used. Experimental infrastructure CAPTIVE logically combines three processes: capture, transform, improve [12].
Comprehensive collection of information integrates methods for identifying patient cohorts based on EHR analysis, containing granular data on individual cases of oncological care, with methods of accumulation in a single database from other sources: databases of randomized clinical trials [13] [14], results of patient questionnaires (in including treatment outcomes reported by patients - PROMs) [15], information from registries [16]. This comprehensive range of information resources allows you to exponentially increase the resolution of each associated semantic level and provide processing in conditions of incomplete data and the presence of noise in digital images of EMC [12].
After integration and fusion, digital information is transformed into factual knowledge using a variety of algorithms, mapping and validation series [17] [18]. The developed extraction technique implies clinical phenotyping [19] based on structured and unstructured data by transforming the results of interaction between patients and healthcare providers into retrospective longitudinal records with the definition of the sample of interest [12]. Custom extractors allow each variable to be absorbed with high precision, using, for example, technologies such as processing natural language variables to populate the production database with information on outcomes from unstructured EHR databases [18]. Optimal algorithms allow accurate identification of clinical outcome records, especially patient-focused information with high computational performance [12].
The ultimate goal of the development of the experimental CAPTIVE system is to study such real-world data as: adherence to clinical guidelines, indicators of quality assessment criteria, the comparative effectiveness of medical technologies and patient-oriented healthcare, support for the adoption of independent doctors [20] [12], as well as joint decisions with the patient [21] [22] [23].
The authors of the publication emphasize that advanced technologies for analyzing EHR records provide ample opportunities for detailed monitoring of the quality of everyday oncological practice and improving control technologies, the effective use of digital data in epidemiological, population and other studies [12]. Obviously, the results of testing analytical systems that have all or some of the properties of the experimental CAPTIVE infrastructure can predetermine the vector of the subsequent innovative development of automated platforms for internal quality control of current oncological practice in real time.

The concept of "smart health" and quality control of oncology practice
The term “smart health” is derived from the concept of “smart planet” [24], put forward by International Business Machines Corporation (IBM) in 2009, and refers to the use of technologies such as portable devices and mobile Internet in order to provide constant dynamic access to information, support of digital communication between all participants in the processes of medical care, etc., which allows you to actively make effective / optimal management decisions, responding to the challenges of the healthcare ecosystem. Information processing in "smart" systems occurs with the help of supercomputers using cloud and / or lake storage of big data and extensive digital networks [24]. The fields of practical application of smart health technologies are directly related to various aspects of quality control of oncological practice and include: digital support of diagnostic processes and selection of treatment algorithms, health management, prevention and risk analysis of oncological diseases, implementation of algorithms for the work of virtual assistants, smart hospitals technologies , clinical and economic research [4].
According to some researchers, “smart health care” is an example of the highest level of evolution of the medical information circuit [4]. Automation of the integration of medical information, of course, accelerates the development of technologies for data analysis in real time, providing fast feedback from doctors in the course of continuous computerized monitoring of the work of medical institutions, emergency network warning and response to violations in the field of quality and safety of oncological practice.

Conclusion
Clinical oncoinformatics has opened a new era in improving the quality and safety of the practice of providing care to patients with malignant neoplasms. It is difficult to imagine the operation of computational and analytical systems outside the automation of continuous feedback processes with medical personnel, as well as the automation of monitoring of mirror (comparative) assessments of control criteria for the quality of medical prevention, diagnosis, treatment, rehabilitation, and social support of patients.
While it is clear that the critical role of big data in improving quality control is clear, interpretation issues remain largely unresolved, as rapid scientific advances in oncology lead to an explosion in a complex variety of characteristics and volumes of variables related to big data. In addition, against the background of the digitalization of healthcare, an increase in the flow of big data exchange, the issues of medical law regarding the protection of personal information, the level of transparency in society and the admission of workers to medical information in oncology, national control and legislative regulation of rights (permits and restrictions) for automated collection, processing and analysis (including by means of artificial intelligence and for the purpose of machine learning), possession (storage), transfer of data and management of electronic resources in the field of clinical oncoinformatics.
The practical value of clinical oncoinformatics methodologies directly depends on the characteristics of operational data, analytical algorithms and properties of high-speed computing hardware and software systems. Integration of all data sources and technical components is a key condition for the development of a system for continuous automated monitoring of the quality of cancer care in real time.

About the authors

Dmitry A. Andreev

Research Institute for Healthcare Organization and Medical Management

Email: AndreevDA@zdrav.mos.ru
ORCID iD: 0000-0003-0745-9474
SPIN-code: 7989-0581

MD, PhD

Russian Federation, Moscow, st. Sharikopodshipnikovskaya, 9

Aleksandr A. Zavyalov

Research Institute for Healthcare Organization and Medical Management

Author for correspondence.
Email: ZavyalovAA3@zdrav.mos.ru
ORCID iD: 0000-0003-1825-1871
SPIN-code: 5087-2394

MD, PhD, Dr. habil., Professor

Russian Federation, 9, Sharikopodshipnikovskaya str., 115088, Moscow

References

Supplementary files