Implementing Big Data Analytic Platform in Healthcare The Israeli experience

doi:10.21203/rs.3.rs-2011150/v1

Background: Medical big-data processing enables analysis of complex multifactorial clinical situations, assessing medical decisions alongside hospital strategic planning and business goals. However, accessing this data is challenging due to legal-ethical, technical and methodological barriers. It also requires the cooperation of multiple partners. Other health systems also struggle to balance scientific innovation and regulations.

Purpose: to establish a practical functional integrative model to overcome these substantial barriers.

Methods: An anonymous big data cloud based data warehouse was created de novo using artificial intelligence algorithm. Major barriers to data access and anonymization were identified and targeted solutions were constructed.

Results: An operating model provided secured anonymous data to ongoing four internal research projects in a single tertiary state medical center. Additional four state medical centers joined the program.

Conclusions: our experience demonstrates the feasibility of creating an integrated functional dynamic medical big data, accessible by multiple users in a virtual cloud. Further studies will determine its cost-effectiveness and potential value for medical research and biomedical industry.

A step by step implementation, involving all relevant stakeholders enables an acceptable national model despite local barriers.

big data

health informatics

regulatory- scientific mechanism

barriers

Appropriate mining and processing of medical big data has emerged as invaluable tool for clinical decision-making. This ever-growing world of data provides clinicians, researchers and pharmaceutical industries with unprecedented capabilities of analyzing complex and multifactorial medical situations. Data sharing also encompasses advantages for hospitals to improve strategic planning and business goals (Vest, 2022), as well as their accountability to care provision (Nwafor, 2022).

However, accessing these constantly accumulating data is challenging due to multitude legal- ethical (anonymizing patient's personal details), technical (data collection, storage and cyber security) and methodological (new modes of analytics and data interpretation) barriers. Adding to these difficulties are the significant differences in local and national regulation of public/state and privately owned health systems. Thus, there is a strong need for functional big data ecosystem capable of providing facilitated access of relevant medical data to potential users being flexible enough to adapt itself to various regulatory systems, ethical restrictions and diverse databases. Obviously, in order to succeed, such a big data analysis platform (BDAP) (Hemingway, 2018) should involve a multidisciplinary partnership composed of health system regulators, data mining experts, clinical researchers and ideally cloud based easy access digital data platform.

Since 2015, and moreover since 2017, a growing interest in big data algorithms for decision making in daily practice emerged in various clinical areas, from cardiology (Hemingway, 2018), through oncology (Resteghini,2018), neuro-ophthalmology (Moss, 2019), psychiatry (Rutledge, 2019) to Covid 19 pandemic (Bragazzi, 2020). Moreover, gathering data from devices expanded opportunities to assess complex patients and detect signs for deterioration to improve critical care (Rush, 2019). Applying new skills and tools in understanding the use of artificial intelligence facilitated research in imaging (Pisano 2019). Practically, each medical field can be observed through the lens of BD algorithms, and it has diffused also to public health issues (Benke, 2018) and health policy, such as strategies for socio-ethical inequity (Zhang, 2017), or suggest a better approach for personalized medicine by a "corrected" gender complexity (Carnevale, 2021). Further implications based on these huge sources of data, were to create predictive models (Obermeyer, 2016).

Accumulative health databases that operate on a national level face two crucial challenges: placing big data in an easily accessible site such as a virtual cloud and by the same time, taking care of patient confidentiality. However, none of the above reports resulted in an integrated functional system on a national level allowing various users such as policy makers and researchers facilitated access to big data while securing patient anonymity.

Here we describe the various steps in the creation of the Israeli national BDAP model from its initiation over a decade ago to its current up-dated state in a tertiary 900 hundreds bed state owned hospital. Technical issues such as engineering data warehouse (Chandra, 2018, Arora, 2011) which are essential steps and should not be ignored are not described here, being beyond the scope of our current paper.

We believe that the experience gathered in this process may provide a sound basis for similar initiatives in other health systems and countries.

Analytic platform design

This article describes the utility of accumulative hospital database as a platform for decision-making and research was a result of a technological and conceptual evolution. Table 1 demonstrates the successive steps of the Israeli Big Data Analytic Platform (BDAP) formation over the last decade (a flowchart of three years joint venture).

Step 1- Initiation: Following initiating "Digital Israel" in 2009 (www.gov.il/en/departments/digital_israel) one of the authors (OT) was involved in the foundation of the national Digital Health project inspired by the need to establish a united national medical record for all citizens. During 2018 a call for innovative projects to encourage research based on this accumulative data was published, meeting the following criteria: (A) a public- private joint venture, (B) support by a start-up company, (C) showing a potential of a long lasting beneficial continuity to professionals and/or the public. We accepted a two-year research grant to fulfill this mission.

Step 2- Design Partnership: to create an effective hospital- start up encounter, a steering committee was established including the hospital deputy director, director of internal medicine ward, head of hospital research forum, director of IRB Helsinki committee, the legal consultant and head of Information Technology (IT) unit. Their role was to establish guidelines for action, project milestones and a contract among the partners. This was followed by an acceptance among all stakeholders: hospital's steering committee, the regulatory and innovation departments of the Ministry of Health (MOH) and the start-up company.

Step 3- Data warehouse: all the hospital medical records were screened to ensure completion of documentation, and accessibility. To certify that these records are not affected by research activity, a replica of the original data was used for further anonymization.

Step 4- Anonymization: A strategy prioritizing better anonymization at the expense of flexible retrieval of individual data was chosen. Therefore, a dual anonymization was performed: a).concealment of demographic details by the hospital cyber team. b). camouflage of all other identifying elements by the startup company.

Step 5- Authorizing and initiating research activities:

5.1 Role of the hospital: a sub-committee was established to assess and regulate research proposals. Criteria for approval were: involvement of an internal researcher, research benefit of big data (vs. epidemiologic) analysis, and approval by the local IRB committee. All proposals received access to the data, unless a conflict of interest and/or potential harm the patient or the organization were identified.

5.2 Data outsourcing: Using a newly established start-up company, all anonymized data was transferred and stored in a "cloud" based "virtual research room", accessible only to institutional authorized predefined researchers for further data analysis. An academic data analyst was recruited and further trained in medical data retrieval and processing by the startup company. A "cloud committee" was established to set guidelines for data storage.

An external cyber consultant validated information security and our preparedness to block potential threats (Kim, 2017). A double verification was conducted to reassure a tight barrier, using simulations of cyber-attacks in hospitals elsewhere (McKeon, 2021). Thus, patient privacy and confidentiality were kept and facilitated access to data was secured.

5.3 Sharing data with additional healthcare organizations: Application was made to additional hospitals to share their data and expand pooling towards a national database. However, we witnessed hesitance among managers to do so, possibly due to uncertainty or reduced trust in the innovative process itself, or the startup company.

5.4 An AI- ML based algorithm was created for data accumulative continuous data flow analyzing trends of services utilization. The algorithm was later adjusted for each research, based on research objective and need.

Step 6- Call for researches: An early call for hospital researchers ("internal path") resulted in three approved proposals by in-house physicians, followed by a forth proposal after 6 months. A successive call for partners from other organizations including other hospitals, academic institutions and industry ("external path") yielded seven proposals.

Step 7- Pilot research projects: To determine the function of the newly established data analysis system four independent researchers received access to the cloud based virtual data room with an ongoing support by a professional data analyst. Three of these research projects successfully completed data mining and processing reaching final analysis and reporting, in spite of some difficulties and barriers (Table 2).

Step 8- Applying an Artificial Intelligence (AI)(3) algorithm and Machine Learning (ML)(4) : following mining the data through a secured cloud, an AI algorithm suitable to each research question was used to set a model and to validate its result.

Step 9- Merging with a Big Data (BD)(5) national research project: In parallel to our initiative and in view of the rising clinical need, the MOH established a group of data engineers to support decision making by hospital managers. Our local project was presented to the MOH, for feedback and continuous funding.

Step 10- towards a nationwide venture: four other state hospitals joined the project, to create a research community utilizing the same mechanism and database. An academic committee prioritized research proposals from these hospitals.

We describe here an ongoing effort on a national level integrating professional and regulatory elements, resulting in the creation of a functional dynamic medical database placed in a virtual cloud, accessible to multiple users.

The 6 "V"s of big data characteristics: value, volume, velocity, variety, veracity and variability (Ristevski, 2018) hold many opportunities for research and therapy, presenting immense potential for improving the quality of care, reducing errors, and lowering the cost of care (Mehta, 2018). On the other hand, it poses challenges of privacy, interpretation, relevant stakeholders collaborate and adapted systems (Pastorino, 2019).

However, the existence of big data accumulation per se does not necessarily indicate its accessibility to a wide range of neither potential users nor its utilization in an analytic manner by the additional layer of AI and ML. The Organization for Economic Co-operation and Development (OECD) required the national electronic health record systems to contribute to national health information and research in a readily manner and that governments will achieve regulatory governance of data privacy, both data from medical records and real would evidence (Eichler, 2019). However most of the researchers refer to a more practical definition, linked to processes such as data collection and data processing (Favaretto 2020). A survey conducted in 2016 revealed only few countries including Norway, Iceland, Finland, and to a lesser degree- UK, USA and Singapore, achieved both readiness to access data and regulations to secure privacy (Moore, 2019). Some countries such as Israel, Spain and Luxemburg have advanced Electronic Medical Record (EMR) (Pisano, 2019), yet partial confirmation for governance of privacy (Eichler, 2019). Others, such as Ireland for example, reached some privacy regulatory mechanisms but not full technical control of EMR (Fraser 2018), while France, Japan and Australia still straggle with both mechanisms to utilize data (Gaze 2003).

Thus, there is currently a wide spread absence of a national integrated and functional modern big data mining and analytical system that provides a flexible, dynamic and sophisticated response in a friendly manner to a wide array of users.

Their main challenges we faced were an insufficient infrastructure and the lack of data polling mechanism. These required experts from commercial companies for support and bridge these gaps. Our clinical staff demonstrated some hesitance and uncertainty to cope with these unfamiliar tools. Therefore, building trust in this innovative environment was essential. Other questions discussed with the Ministry of Health were the strategy to update data of an ongoing research, pooling data from many medical centers to expand our database, and shared responsibility for data mining, security and processing between the hospital and MOH (8). Legal issues such as intellectual property should also be considered.

Several features make our model unique. It is a novel joint venture integrating diverse partners including a state regulator (MOH), public medical facility (governmental tertiary state hospital) and emerging Startup Company. It offers a comprehensive research oriented service including not only the database per se, but also the additional dynamic data mining and analytic services which are crucial for all research purposes. Furthermore, the facilitated access to the virtual cloud containing data saves valuable time to all authorized researches by minimizing the need of going through time-consuming regulatory beaurocracy. Together, these elements commonly grouped as a functional model provide a sound basis for its expected future success. These expectations should be met with caution. The absence of guidelines, shortage of trained IT professionals, lack of resources and slow regulatory responsiveness were challenging and considerably delayed the process. In addition, the understandable reluctance of researchers to exploit new pathways of big data analysis, as well as the hesitancy of colleagues to share data, are expected difficulties in the implementation of this model. Moreover, one should consider the potential threats by external forces, such as COVID 19 pandemic, that may shift the focus of efforts and tendency to deal with academic challenges.

Whether this is an optimal model of big data analytic platform in healthcare, which can be generalized to other systems, remains an open question. Obviously, various legal aspects of privacy and contracts should be carefully adapted to local national, academic and hopefully industrial environments. Nevertheless, big data analytics and its associated need for efficacious processing are ever increasing in health and many other fields, and we can no longer ignore this trend. Scientists, clinical researchers and hospital staff should encouraged by their national and local leadership via focused education and training to take fruitful advantage of these new avenues.

The Israeli experience demonstrates the feasibility of creating a functional dynamic medical big database, accessible to multiple users in a virtual cloud. A step by step implementation, involving all relevant stakeholders enables an acceptable national model despite local barriers. Further long-term real life studies will determine its cost, effectiveness and academic impact.

Ethics approval and consent to participate - Not applicable: the research deals only with a mechanism, no ethics approval was required
Consent for publication- Not applicable. Both authors approved the publication
Availability of data and materials- The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Competing interests- The authors declare that they have no competing interests and no conflict of interest.
Funding- no funding was allocated to conduct the research. It was a pilot project of the KINERET big data national project initiated by the Israeli Ministry of Health, Division of governmental medical centers.
Authors' contributions- both authors were involved in describing the process, analyzing barriers, writing the manuscript and approving the final version.
Acknowledgements- we thank the Division of governmental medical centers for promoting big data research.

Arora, M. Gosain, A. (2011). Schema Evolution for Data Warehouse: A Survey. International Journal of Computer Applications 22, 6-14.
Benke, K. Benke, G. (2018). Artificial intelligence and big data in public health. Int J Environ Res Public Health. 10, 15(12).
Bragazzi, NL. Dai, H. Damiani, G. Behzadifar, M. Martini, M. Wu, J. (2020). How Big Data and Artificial Intelligence Can Help Better Manage the COVID-19 Pandemic. Int J Environ Res Public Health. 2, 17(9).
Carnevale, A. Tangari, EA. Iannone, A. Sartini, E. (2021). Will Big Data and personalized medicine do the gender dimension justice? AI Soc. 1, 1–13.
Chandra, P., Gupta, MK. (2018). Comprehensive survey on data warehousing research. Int j inf tecnol. 10, 217–224
Eichler, HG. Bloechl-Daum, B. Broich, K. Kyrle, PA. Oderkirk ,J. Rasi, G. et al. (2019). Data rich, information poor: can we use electronic health records to create a learning healthcare system for pharmaceuticals? Clin Pharmacol Ther. 105(4), 912–922.
Favaretto, M. De Clercqm E. Schneble, CO. Elger, BS. (2020). What is your definition of Big Data? Researchers’ understanding of the phenomenon of the decade. PLoS ONE. 25, 15(2), e0228987.
Fraser, AG. Butchart, EG. Szymański, P. Caiani, EG. Crosby, S. Kearney, P. et al. (2018). The need for transparency of clinical evidence for medical devices in Europe. Lancet. 392, (10146), 521–530.
Gaze, B. (2003). Privacy and research involving humans. J Law Med. 10(4), 410–434.
Hemingway, H. Asselbergs, FW. Danesh, J. Dobson, R. Maniadakis, N. Maggioni, A. et al (2018). Big data from electronic health records for early and late translational cardiovascular research: challenges and potential. Eur Heart J, 39(16), 1481–95.
Kim, S. Lee, H. Chung, YD. (2017). Privacy-preserving data cube for electronic medical records: An experimental evaluation. Int J Med Inform. 97, 33–42.
McKeon, J. Lawsuit Links Baby Death to AL Healthcare Ransomware Attack. [cited 2021 Oct 3]; Available from: https://healthitsecurity.com/news/lawsuit-links-baby-death-to-al-healthcare-ransomware-attack
Mehta, N. Pandit, A. (2018). Concurrence of big data analytics and healthcare: A systematic review. Int J Med Inform. 114, 57–65.
Moore, W. Frye, S. (2019). Review of HIPAA, part 1: history, protected health information, and privacy and security rules. J Nucl Med Technol. 47(4), 269–272.
Moss, HE, Joslin, CE. Rubin, DS. Roth, S. (2019). Big Data Research in Neuro-Ophthalmology: Promises and Pitfalls. J Neuroophthalmol. 39(4), 480–486.
Nwafor, O. Johnson NA. (2022). The effect of participation in accountable care organization on electronic health information exchange practices in U.S. hospitals. Health Care Management. Review. 47(3), 199-207.
Obermeyer, Z. Emanuel, EJ. (2016). Predicting the Future - Big Data, Machine Learning, and Clinical Medicine. N Engl J Med. 29, 375(13), 1216–1219.
OECD. (2015). Health data governance: privacy, monitoring and research.
Pastorino, R. De Vito, C. Migliara, G. Glocker, K. Binenbaum, I, Ricciardi, W. et al. (2019). Benefits and challenges of Big Data in healthcare: an overview of the European initiatives. Eur J Public Health. 29, (Supplement_3), 23–27.
Pisano, ED. Garnett, LR. (2019). Big data and radiology research. J Am Coll Radiol. 16,(9 PundefinedB), 1347–1350.
Resteghini, C. Trama, A/ Borgonovi, E. Hosni. H. Corrao, G. Orlandi, E. et al (2018). Big data in head and neck cancer. Curr Treat Options Oncol. 25;19(12):62.
Ristevski, B. Chen, M. (2018). Big data analytics in medicine and healthcare. J Integr Bioinform. 10, 15(3).
Rush, B. Celi, LA. Stone, DJ. (2019). Applying machine learning to continuously monitored physiological data. J Clin Monit Comput. 33(5), 887–893.
Rutledge, RB. Chekroudm, AM. Huys, QJ. (2019). Machine learning and big data in psychiatry: toward clinical applications. Curr Opin Neurobiol.15,55,152–159.
The National Digital Program. [cited 2021 Jul 5]Headquarters for the National Digital Israel Initiative.; Available from: https://www.gov.il/en/departments/digital_israel
Vest, JR. Freedman, S. Unruh, MA, Bako. AT, Simon, K. (2022). Strategic use of health information exchange and market share, payer mix, and operating margins, Health Care Management Review, 47(1), 28-36
Zhang, X. Pérez-Stable, EJ. Bourne, PE. Peprah, E. Duru, OK. Breen, N. et al. (2017). Big data science: opportunities and challenges to address minority health and health disparities in the 21st century. Ethn Dis. 20, 27(2), 95–106.

Table 1 Flowchart- Our Timeline- Actions and Challenges (November 2019 to May 2022)

Time line	Step	Action	Status/ challenges
July 2018-	Step 1	Initiation	Completed by schedule
August 2018	Step 2	Partnership	Established
February 2019	Step 3	Data warehouse	Completed in-house
May 2019	Step 4	Anonymization	An external consulted to validated methodology
October 2019- March 2020	Step 5	Authorizing and initiating research activities	A need to establish guidelines, regulator acceptance
June 2020	Step 6	Call for researches: "internal / external path"	Hospital physicians were engaged with COVID-19 pandemic
December 2020- February 2021	Step 7	Pilot research projects	Initiated *
March- May 2021	Step 8	Applying an AI algorithm and ML	Completed by the start-up and MOH data scientists
July- October 2021	Step 9	Merging with a national project to initiate BD research	Lack of infrastructure and data polling mechanism
December 2021- May 2022	Step 10	Creating a state hospitals research community	Establish a mechanism to recruit other hospitals to share data
*Active researches are described in table 2.

Table 2 Research Projects- Summary/ Status of Proposal Applications by In-House Physicians (Internal Path)

	Researcher characteristics	Main topic	Status	Barriers
				Early phase	Late phase
1	MR physician, Head of Internal Medicine Department	A predictive model of patients flow & occupancy in Internal Medicine wards	Completed	Creating the mathematical algorithm	Leadership hesitancy to adopt conclusions
2	MB physician, Head of Clinical Pharmacology	Analyzing drug side effects in hospitalized patients	Delayed	Partnership with another medical center	-
3	OT physician, Medical Deputy Director	Predicting COVID 19 patients medical utilization & hospitalization trends	Completed (analysis in process)	Regulatory approval for database	-
4	SM physician, Head of Cardiac Catheterization Unit	Clinical prediction of successful catheterization technology to improve survival	Completed (analysis in process)	Lack of an Analytic tool	Lack of skilled personal

No competing interests reported.

Implementing Big Data Analytic Platform in Healthcare The Israeli experience

Status:

Version 1

Abstract

Introduction

Results

Analytic platform design

Discussion

Conclusions

Declarations

References

Tables

Additional Declarations

Status:

Version 1