Innovation and technology are driving efficiency and performance in clinical research by changing the way data are captured, monitored, analyzed, and reported. Mobile and Internet-connected medical devices generate large volumes of data in an environment where data science, data discovery, and visual analytical tools are empowering clinical researchers and study teams to improve trial management, trial monitoring, and trial performance.
Over the years, contract research organizations (CROs) have built multiple applications using a variety of technologies to address the unique needs of their customers. Legacy applications lack integration, lack access to a single source of truth, and do not provide actionable information to improve the efficiency of trial conduct. Disparate systems containing redundant and inconsistent data are a major challenge for study teams trying to make the right decision at the right time on various business, operational, and clinical issues.
In addition to legacy application challenges, the complexity of studies can dramatically increase with the introduction of the “medical Internet of things” (MIOT)—an ecosystem of integrated infrastructure connecting people, processes, data, and medical devices to capture and process health-related data or to provide health-related services. MIOT devices generate a large volume of data, and some of the critical datasets need to be processed in real time.
Real-time data capture and validation reduces manual data entry errors and improves overall data quality and accuracy. Though wearable devices and other automated forms of data capture technologies are improving, they still lack integrated workflow to extract structured and unstructured data and to validate and transform data into a meaningful form for further analysis.
This paper discusses clinical trial challenges and the need for analytical tools, and proposes an integrated clinical analytics model to improve business and operational excellence.
Clinical Trial Challenges
The high cost of research and numerous rules and regulations put the drug development industry under pressure to conduct clinical trials effectively and efficiently. A clinical trial is a scientific process leveraging other practices, including project management, manufacturing, and supply chain management. A recent report from Battelle states that “In 2013, the biopharmaceutical industry alone invested $10 billion in clinical trials, with a total reported 1.1 million patients enrolled—or approximately $9,090.91 spent per patient.”1
Because of a high level of investment, compliance needs, safety concerns, and security guidelines, organizations depend on data-driven analytics to improve decision-making and prevent potential future issues.
Need For Analytics
An analytics solution provides unique insights, generates new knowledge, and improves outcomes. To optimize performance and improve decision-making, data-driven analytics should focus on standardization of datasets.2 Organizations use analytics for business improvement, cost reduction, and customer experience improvement. For example, by combining analytics with manual efforts, one health system could reduce audit expenses by 75%.3
Traditionally, data managers and study teams identify data quality issues by analyzing the data during certain milestones or at the close of the study. This approach delays the process of addressing data quality, safety, and other critical issues to later in the study. Analytics combined with machine learning could allow for better insights and improved predictability.3
A 1999 report by the Institute of Medicine (now known as the Health and Medicine Division of the National Academies of Sciences, Engineering, and Medicine) stated that medical errors each year take about 98,000 patient lives and cost hospitals nearly $29 billion.4 This report also cited that three out of four errors could be eliminated by centralizing and integrating information, and by improving the availability of information about drugs and patients when needed.
Analytics are used for predicting patient health and for gaining insights on related areas like fraud detection, communication, and education. However, gathering and processing data for analytics poses challenges that are both qualitative and quantitative.3
Organizations are collecting more data about patients, and the rate of collection increases with personalized and MIOT devices. These data could be used to find simple and complex relationships, and the analysis could help to improve patient care, to prevent medical malpractice, to increase healthcare efficiency, and to support insurance and payments. Further, advances in technology and systems have generated a large volume of health data.5 Analytics solutions empower healthcare professionals to improve clinical decision-making, predict risks, monitor patients, and manage finances.
Data Types and Storage
Visual analytics use interactive visual interfaces to gain insights from complex and large datasets that are processed on a near real-time basis. This can also be used to build dashboards for monitoring various health factors, guidelines, and compliance.3 However, organizational data are going to be distributed and stored using different types of technologies. Further, data entities are represented using a variety of standards that may not be consistent with unified reporting and analytics.
As shown in Figure 1*, clinical research data can be classified into a variety of functional areas in terms of their foundational, transactional, operational analytics, clinical analytics/study reporting, and business analytics purposes. In addition to the sources of data in these areas noted in Figure 1*, some study leaders may also use external databases, social media, and other market sources to compare their data to data from the overall industry or specific competitors.
The following five different types of analytics may be used in clinical research:
- Transactional Analytics: Clinical trials depend on many different types of applications, and on interactions with various internal and external users to capture and process data in real time. Transactional analytics focus on data at the transaction level; this type of analytics processes records in real time to improve specific outcomes. Clinical data managers can leverage transactional analytics to gain deeper insights on transactional data to improve quality by applying consistent business rules and policies.
- Operational Analytics: This is the next level of transformation from traditional business intelligence. Operational analytics is a complex analytics process that consolidates various operational data sources to provide insights on current operations.6 Clinical research associates, clinical project managers, and principal investigators can use operational analytics to support decision-making and improve implementation and monitoring of clinical trials. This includes application to such tasks as site selection, monitoring site performance, patient recruitment, drug distribution, managing payments, and scheduling.
- Clinical Analytics: This helps researchers to compare current clinical study data to those from similar studies conducted internally based on therapeutic or domain-specific research. In addition, current data could be compared to past clinical trials data to predict or improve safety, efficiency, and efficacy of new medicines.
- Predictive Analytics: Implementing predictive analytics requires current and past data related to studies conducted within an organization, plus additional relevant data from external and industry sources to model and predict certain types of events. Predictive analytics is a valuable tool for researchers to run trials effectively and improve key aspects of clinical research by unifying current and historical data to predict future events, prevent failures, and prescribe certain actions.
- Business Analytics: This helps to run the business operations efficiently by using available time, budget, and resources. It also facilitates decision-making to improve and grow the business by analyzing challenges and opportunities in managing clinical trials. This includes gathering insights on Time, Cost, Scope, Human Resources, Stakeholders, Quality, and Risk Management. Business analytics could also be used for portfolio analysis by grouping together a similar set of related projects and focusing on high-risk project and related metrics.
Integrated Clinical Analytics Model
The use of analytics leverages existing organization data at various levels and provides value and insights for better operational planning and management. It requires an understanding of current operations and uses insights to improve performance and efficiency. To implement an analytics solution, organizations should create a consistent framework and take into account organization culture and policies. Further, analytics programs should be implemented incrementally, to avoid any potential disruptions to existing operations and customer impact.7
The proposed integrated clinical analytics model (ICAM) shown in Figure 2* will segment data analysis into multiple levels, with each level having a different purpose, scope, and focus area. The goal is to “learn fast and learn often” on various business, operational, and clinical functions.
The framework of this model is to learn using analytics, transform learnings to build a knowledge base, and use the knowledge base to create business rules and policies to govern and integrate with existing systems and processes.
The proposed ICAM will directly help clinical project managers, clinical research associates, study coordinators, and principal investigators measure, analyze, and improve study outcomes. Some of the applications are:
- Improve site feasibility and site selection based on past performance
- Improve site monitoring rules based on specific risks
•Improve patient engagement based on current performance, behavior, etc.
- Improve data quality on based on trends, control limits, etc.
- Improve supply chain logistics for drug distribution
Implementation of this model involves dividing analytics solutions into multiple levels based on focal areas, systematically transforming learnings to build a knowledge base, and creating a workflow to apply the learnings to augment existing business rules, policies, checks, and audits.
Business processes are highly complex, and many organizations depend on information technology (IT) to identify, define, and manage rules digitally to achieve business goals and objectives. This includes defining and implementing business boundaries, business rules, policies, and regulations to conduct business efficiently.
A business rule is a means of managing various types of business domains and their components; it is a set of defined activities, rules, and constraints integrated with the business process workflow.8,9 To gain insights on patient information, data residing in various places and different types of formats need to be consolidated into one repository.10 This will help to predict and avoid various risk factors, including patients’ compliance with appointments and medication schedules, and other behaviors during participation in trials.
Data preparation is one of the foundational processes in implementing any analytics solution. The first step is building a standard data dictionary for analytics by analyzing existing data sources. This is a collaborative effort undertaken with data stewards from different business functions, data architects, and development teams. Ideally, this effort should focus on a so-called “single source of truth,” and avoid using secondary data which might have been transformed for other purposes or modified for individual use.
The second step is data mapping from source to target by using associated transformation rules. This is a critical step of the process to ensure that consistency is applied for similar datasets and that all business rules for transformation are consolidated into one place for review and applied for future changes.
The final step is grouping data to build a unified dataset based on the type of analytics and audience. One way of grouping data is based on business, operational, and clinical focus areas. The next section discusses common challenges of implementing an analytics solution.
Analytics implementation is a transformational program impacting people, process, and technology. As described below, implementation faces challenges relating to the areas of Data, Privacy, Security, Validation, Governance, Training and Development, and Outcome.
Data: Creating a unified data analytics model is a complex process. The widespread distribution of data using multiple technologies, along with nonstandard and unstructured data, poses technical and process challenges in implementation. Communicating overall vision, strategic objectives, and business outcomes to all stakeholders and building a business-IT partnership will help to reduce resistance and build support to standardize datasets owned by various business functions of the organization.
Privacy: Analytics provide wider connectivity and deeper insight on multiple data sources. Users can filter from large datasets and narrow down to specific data, which may expose personal identity or provide more information of a patient taking a specific type of treatment. Organizations should implement governance and review process on data standardization, data access, and controls on privacy-related exposure and risks.
Security: Analytics solutions should be carefully evaluated for integrated identity and access management policies to ensure that authentication and authorization are consistent with the rest of the systems in the organization. If the analytics solution includes data from electronic health records and personal health information, fine-grained access control needs to be implemented to govern and control access to these sensitive data. The system should be built with sufficient auditing and alerts to prevent any improper or fraudulent use of personal information beyond the intended use.
Validation: Modern analytics tools are shifting from IT-led to business-led solutions. Proper governance and validation on data and processes should be built to ensure consistency, quality, and accuracy. Scripts to extract, transform, and load data should be validated for compliance with 21 CFR Part 11 (Electronic Records; Electronic Signatures) of the Code of Federal Regulations on data extraction, data transformation, system and change controls, and other guidelines.
Governance: Analytics empowers business leaders to perform their own analyses by unifying diverse sets of business and operational data. Business teams need to work closely with IT to understand data source, usage, security, and access. This will ensure that a consistent process is followed from data extraction, transformation, and exposure. In addition, proper governance should be in place for adding new data sources and allowing audit trails, for sharing data, and for generating custom dashboards.
Training and Development: Analytics teams need skilled resources with a good understanding of business and industry, data sources, data entities, and data usage. Having a well-trained workforce is one of the challenges organizations face while implementing big data analytics solutions.11 The organization should perform a good skills assessment and fill the resource gaps with training or by recruiting external sources with experience in executing similar projects.
Outcome: Creating an analytics solution requires a high level of investment, and implementation may run from many months to years. Executing projects with agility by dividing large requirements into small chunks will help business teams to understand the need, measure the outcome, and realize the incremental value of the project goals and objectives.
Recommendations and Discussions
This paper provides an analytics model, approach, and implementation strategy to transform learnings and insights to improve productivity and performance. Implementation of this model requires identifying the unique needs of an organization and understanding its current processes, data, and systems. Empowering business with a highly sophisticated tool without proper governance may result in multiple dashboards, redundant and inconsistent datasets, uncontrolled data access, and unintended privacy and security issues. For successful implementation, organizations should create a sustainable review and governance process, including paying attention to data governance, data standards, data security, and extensive training on various tools and technologies used in analytics.
Modern clinical research depends on analytics to improve decision-making and gain insights on business and operational performance. This article has discussed the challenges and needs surrounding analytics in clinical research, looked at various analytics types, and presented the ICAM to transform insights gained by individuals into institutional knowledge. The author also has discussed an analytics implementation strategy and various implementation challenges in the clinical research environment. The proposed model provided a framework to generalize, transform, and apply analytics insights to improve business and operational excellence.
The author thanks Brook White, executive director of corporate communications for Rho, Inc., for reviewing this paper and providing valuable feedback and comments.
- S Nair, LA Celi. 2017. Big data applications in clinical medicine. Bus Intell J, 22(1):19–25.
- BI Reiner. 2013. Creating accountability in image quality analysis—part 4: quality analytics. J Dig Imgng, 26(5):825–9. doi:http://dx.doi.org.libdatab.strayer.edu/10.1007/s10278-013-9628-1
- AF Simpao, LM Ahumada, JA Gálvez, MA Rehman. 2014. A review of analytics and clinical informatics in health care. J Med Systs, 38(4):1–45. doi: http://dx.doi.org.libdatab.strayer.edu/10.1007/s10916-014-0045-x
- S Venkatraman, H Bala, V Venkatesh, J Bates. 2008. Six strategies for electronic medical records systems. Comms ACM, 51(11):140–4. doi:10.1145/1400214. 1400243
- AF Simpao, LM Ahumada, MA Rehman. 2015. Big data and visual analytics in anaesthesia and health care. Brit J Ansths, 115(3):350–6. doi:10.1093/bja/aeu552
- PR Newswire press release. 2016. Research and markets—operational analytics market to reach $10 billion by 2021—key innovators are Alteryx, Cloudera & Evolven Software.
- HealthInfo. 2016. Analyticsdriven operational performance improvement represents enormous economic opportunity for hospitals. Hlthcr Infrmtcs, 33(6):16–7.
- C Paolo, A Antonia, D Ernesto, L Mariangela, M Manuela, C Angelo. 2016. Translating process mining results into intelligible business information. Prcdngs 11th Intl Knwldg Mgmt Orgs Conf (KMO ‘16). ACM, New York, N.Y. Article 14. doi:http://dx.doi.org/10.1145/2925995.2925997
- B Jose, F Paolo, M Maurizio, D Jia. 2010. Dynamic context-aware business process: a rule-based approach supported by pattern identification. Prcdngs 2010 ACM Symp Appl Cmptng (SAC ‘10). ACM, New York, N.Y. 470–4. doi=http://dx.doi.org/10.1145/1774088.1774186
- JM Buell. 2016. The beauty of predictive analytics leveraging data into action. Hlthcr Exec, 31(5):10–8.
- J Upton. 2017. Big data: is it crunch time for pharma? Pharm Exec, 37(3):12–9.
Kaali Dass, PMP, PhD, (email@example.com) is an information technology (IT) consultant at Rho, Inc. focusing on enterprise architecture, IT strategy, and planning. He also serves as vice president of IT and communications for the North Carolina Project Management Institute in Raleigh, and as an adjunct professor teaching IT and project management for Strayer University.
*To see all figures and/or tables published originally in this article, please visit the full-issue PDF of the June 2017 Clinical Researcher.