The healthcare industry’s quest to use and build data-driven organizations has faced two major challenges: 1) finding the right candidates with the skills to do the work, and 2) creating effective organizations. Ineffective organizations lack clear structure and role definition for employees, and are slow to adapt to new tools. To be an effective and dynamic organization, leadership support incorporates data resources and metrics developed by stakeholders at the unit/division level to drive organization decisions. The ultimate goal of the data-driven business is to align analytics into the strategic planning process and have business leaders use data to drive decisions. This allows leaders to identify areas of opportunity to improve outcomes for both the organization’s return on investment (ROI) and optimal patient care metrics and workflow.
1. Making data key to driving decisions
There are three major types of organizational structures: functional, divisional, and matrix-ed.
A functional structure divides the organization based on specialty, i.e., people who do similar tasks are grouped together. The advantage of a functional structure is that individuals are dedicated to a single function, can learn from and support each other on tasks, and offer more rapid decision making due to intra-departmental cohesion. Additionally, clearly defined roles and expectations limit confusion. The downside is that it is challenging to facilitate strong communication between different departments, and these teams often work in silos and are not actively involved or understand the work of other divisions within the organization. Due to these silos, functional organizations have a difficult time collecting and using data for ROI, making opportunities to improve performance limited. In a divisional structure, organizations structure leadership according to different products, project, lines of business, or brands. Matrix-ed organizations are a combination of the two. There are less linear. For example, employees have multiple bosses and reporting lines, and not only do they report to a divisional manager, but they also typically have project managers for specific projects. In the case of big data, matrix-ed organizations have a group for information technology, and teams specifically for business intelligence, analytics, and/or big data. All teams work together across multiple projects in an organization.
There are four steps shown to improve organizational ROI using data-driven results.
1) Embed analytics teams at the unit level. Distributed analytics teams who are embedded in each unit are important to not only collect the right information but use it to help improve the organization.
2) Create cross functional teams that have a variety of skills and knowledge necessary to understand what metrics to identify. Building stakeholder teams within the organization to include champions from each division, unit, or project is also vital to successfully identifying key trends and metrics within big data.
3) Align these metrics to the objectives of the organization. Big data is meaningless if it is not tied to the objectives/goals of the organization/division/unit/project. Aligning the right metrics or trends to organizational strategy will improve visibility of issues and factors that require additional resources, research, or monitoring.
4) Develop a charter with leadership buy-in and engagement that supports the review of resources, needs, and tools needed in order for the organization to implement a big data initiative tied to ROI.
2. Set objectives and target data for use in ROI
As with any strategic initiative, a dive into big data should be accompanied with values and measures of success. Value, which could be measured as ROI, is often what drives financial investment in a large project. It is also a measure of success that organizational leadership can hold project stakeholders accountable for meeting financial targets.
ROI is often difficult to quantify in data and analytic projects. In general, assessment in three categories help support ROI determination in a big data project: revenue increase, cost reduction, and cost avoidance. Within healthcare environments, a fourth category – clinical outcomes – can also be measured.
1) Revenue increase – Will new data types support initiatives that were not previously possible using traditional data warehousing technologies? Can new/improved analytic models support revenue growth? Is there a market advantage that can translate to new business or a boost in existing business?
2) Cost reduction – Can existing technologies be retired? How much time can be saved in future projects through the implementation of a big data program? Can productivity be improved?
3) Cost avoidance – Are there known, future expenses that can be eliminated? Does the big data project mitigate risks that could have costs if they occur?
4) Clinical impact – How many lives are impacted? How can clinical quality outcomes improve? Qualitative value helps to support the big data business case but may not be enough to justify the project and ongoing costs by itself. Often referred to as “soft ROI,” these can include employee (e.g., data scientist) satisfaction, easier access to data, and better information about data (e.g., metadata, lineage).
Role of education
It is critical to work with business units early in the process to educate and socialize the big data project. The initiative may start in the IT department, but it should be driven by the business units, and include their commitment and engagement throughout. Unfortunately, business leaders do not always understand what big data is and why it is different from the data environment in which they are already working. Educational sessions that focus on non-technical topics and practical, real-world use cases can help bring business-minded leaders up to speed. During educational sessions, project sponsors should collect ideas on how the big data platform can solve issues that the business units are facing. These solutions may naturally present new value opportunities.
Examples of measurable success criteria “quick wins”
With the right business use cases and value drivers, measurable success criteria should be relatively easy to develop. Some of the quick wins might be to consolidate disparate data sources into a single platform for ease of use by data scientists and analysts. This can drive down the time to develop and maintain models or other analytic solutions. Process automation can be measured by hours saved. Real-time data streaming, such as HL7 ADTs (Admissions, Discharges, and Transfers), can decrease clinical outreach time for patients who are discharged from the emergency department (ED). Natural language processing (NLP) can extract insights from unstructured clinical data to uncover previously undiagnosed conditions. Storage of legacy application databases can eliminate licensing costs that come with the original transactional software while maintaining the benefit of access to the data for reporting and analytics. Regardless of the incremental value opportunities, project stakeholders should clearly understand what will be delivered, when, and by whom.
3. Defining roles
The number of roles and specialties within the team can vary greatly depending on the size and needs of the organization. While team members may have very specific roles, specializing in distinct areas of focus, it is also possible to find individuals with cross-role expertise. When working to build out your team, it is important to fulfill the workflow segments (data collection, transformation, selection, discovery and insights, and deployment) based on the size and focus of your organization. Figure 1 models the structure necessary to fulfill each of these segments, as well as the usual job titles or role that usually specialize in that area.
Skill sets – hard skills
Within each role in the big data team, there are various skills that will enhance the organization’s
ability to find, process, and implement data use. Each role will have their area of expertise, but in
order to successfully implement change in the organization, the team should be composed of
individuals that are flexible and dynamic with diverse experiences that enhance their ability to
work together. Below in Chart 1 is a list of basic skill sets needed for each of the roles within the
big data team.
There are a variety of tools that are required for each of the tasks. While they are segmented by role for simplicity, many can and will be used across skill sets. New tools and technologies continue to be developed, making it almost impossible to capture all of the means currently in use. Chart 2 shows a simplified view of tools and technologies used by each of the skill sets.1
1 These are commonly used tools, and listed for example purpose only. HIMSS does not endorse use of any particular tools or resources.
Technologies are fast changing, so having a team with diverse skills is key. Being trained the latest “X” is only half as important as individuals that that actively seek continued education and training on a variety of subjects.
Skills sets – soft skills
In order to be effective and successful, data-team members also need to work collaboratively and collectively using a service level approach to achieve deployment of end user tools and integration of change into organizations. It is imperative that each member of the team reaches out across the organization to seek clarification and understanding of needs or requirements, while also ensuring that models are received positively and are being integrated into processes.
Does healthcare experience matter?
Having staff with healthcare experience is most helpful to large institutions in guiding the team, however just having some team members with clinical and/or healthcare expertise is sufficient. It is often easier to teach the use of tools rather than increase healthcare knowledge. In smaller institutions, where team members carry multiple roles, the understanding of healthcare and business analytics becomes imperative.
Bringing it all together
Figure 2 identifies key traits for big data team members in a successful team.
4. Develop baseline skills necessary for each resource at each level
There are four main actions that businesses and organizations need to take for their organizations to develop their resources. These include building a charter, using this charter to create their team, developing a global education plan to train staff, and providing orientation and mentoring to help support the team. Each of these is further outlined below:
1) Charter – As with any project, in order to gain support and funding for the necessary resources and tools, it is imperative that leadership develops goals and objectives for each department or project in a big data charter.
2) Team creation – Using the charter as a guide, the human resources department and departmental champions need to review existing resources for any individuals who have the capacity to assist with the project. New job descriptions should be created and filled with any resource gaps.
3) Global education plan – In parallel to advanced training for existing data staff that might be needed, basic training should be provided to all team members (executives to technicians) so that everyone is informed, enabled, and empowered to suggest ways to use the existing data and tools to improve clinical, financial, and operational practices. This training should include improving familiarity with available data sets, tools, and techniques, as well as conceptualizing and identifying new opportunities from examples sourced from successes within the organization or externally.
4) Orientation and mentoring – By establishing both an orientation and a mentoring program/process, resources will be able to integrate into the team while supporting data ROI objectives/goals.
5. Develop mentoring of roles within the organization
Formal mentoring of each resource is recommended for both resources already working for the organization as well as for those who are just joining. The benefits of mentoring include:
Reduced attrition of resources
Improved commitment to objectives/goals of the organization
Improved quality of data to support objectives/goals Improved career path for data resources
Supports staff career growth
Basic skills are insufficient for staff who have limited experience in an organization to understand what questions to ask, where to get data, how to document and evaluate data, what level of quality is expected in their work product, and so forth. This is why mentoring within the organization is an important part of developing and further honing the cross-functional big data development team.
6. Evaluate results within agreed timeframes
A big data project comes full circle when the sponsor or champion can compare results (i.e., value) to goals and metrics. Whether quantitative or qualitative, it is important to demonstrate that the project accomplished what it set out to do. Some project aspects may still require additional time, however, definitive milestones should show incremental value as early as possible. The following is a short list of evaluation questions to help establish value and ROI:
Sample evaluation questions
Were Extract Transfer Load (ETL) processes redesigned to save time?
Are licensing costs eliminated by archiving data?
Are data that were not available before now usable in analytics? What was the impact? For example:
7. Share results to further opportunities and success
As business units realize value from a big data project, successes should be socialized and shared across the organization. Many parts of the organization may still not comprehend what big data is and what it can do for them. By communicating with business units and sharing the ways in which big data has improved clinical, financial, or operational metrics, big data evangelists can spark ideas for new projects and new value. Robust training and collaboration meetings will keep analysts and data scientists engaged and eager to share their own successes. The more the organizations are aware of what is possible with big data and how they can leverage it, the greater the chances that big data will be a long term, successful part of healthcare.
Resources for Further Learnings
Burgess, Jan. “How to Keep Staff up to Snuff with Big Data Training.” Ingram Microadvisor. Posted November 24, 2018. https://www.ingrammicroadvisor.com/data-center/how-to-keep-staff-up-to-snuff-with-big-data-training.
Cao, Longbing. “Data Science: Challenges and Directions.” Communications of the ACM, Vol. 60 No. 8, Pages 59-68. August, 2017. https://cacm.acm.org/magazines/2017/8/219605-data-science/fulltext.
Currin, Chuck. “Building your Big Data Analytics Staff.” Posted April 7, 2016. https://tdwi.org/articles/2016/04/07/building-your-big-data-analytics-staff.aspx.
Data Science and Cognitive Computing Courses by Cognitive Class AI. https://cognitiveclass.ai/.
Eaton, Charlotte. “Skilled Staff are Key to Making the Most of Big Data.” Supply Management. Opinion Piece Posted February 2015. https://www.cips.org/supply-management/opinion/2015/february/skilled-staff-are-key-to-making-the-most-of-big-data/.
Elite Data Science. “Free Data Sources for Beginners.” May, 2017. https://elitedatascience.com/data-science-resources.
Hitchcock, Erin. “5 Big Data Job Descriptions to Hire an All-Star Team.” Datameer. Blog Post Posted February 27, 2018. https://www.datameer.com/company/datameer-blog/big-data-job-descriptions-hire-recruit-team/.
Kelly Mitchell. “Scientists, Hygienists, Explorers: Three Essential Big Data Crew Members.” Blog on Big Data. Posted March 27, 2014. http://kellymitchell.com/2014/03/scientists-hygienists-explorers-three-essential-big-data-crew-members/.
Macsin, Andrea. “Stop Hiring Data Scientists if You’re Not Read for Data Science.” Data Science Central, Guest Blog posted July 31, 2015. http://www.talentanalytics.com/blog/stop-hiring-data-scientists-if-youre-not-ready-for-data-science/.
Olavsrud, Thor. “How to Close the Big Data Skills Gap by Training your IT Staff.” CIO from IDG. Posted October 2, 2013. https://www.cio.com/article/2382064/big-data/how-to-close-the-big-data-skills-gap-by-training-your-it-staff.html.
Sussman, Jason. “Three Ways Health-Care Leaders are Leveraging Big Data.” CFO An Argyle Company. Posted May 26, 2017. http://ww2.cfo.com/health-benefits/2017/05/three-ways-health-care-leaders-leveraging-big-data/.
This paper is a product of the HIMSS C&BI Committee and its Task Forces.
Ellen Harper, DNP, RN-BC, MBA, FAAN
Clinical Assistant Professor
University of Kansas School of Nursing
Claudine Beron, PMP
Initiate Government Solutions, LLC
Work Group Lead
Amber Zimmermann, MBA, RN-C
Sr. Manager, Health Data Analytics
Amber Zimmermann, MBA, RN-C
Sr. Manager, Health Data Analytics
Senior Director - Informatics
Claudine Beron, PMP
Initiate Government Solutions, LLC
Shelley Price, MS, FHIMSS
Director, Payer & Life Sciences
Joanne Bartley, CAE
Manager, Health Business Solutions
Source: HIMSS C&BI Committee
by Amber Zimmermann, MBA, RN-C Sr. Manager, Health Data Analytics | Royal PhilipsClaudine Beron, PMP CEO | Initiate Government Solutions, LLCMitch Kwiatkowski, Senior Director, Informatics | Gateway Health Plan
This resource helps executive leaders and IT/informatics/clinical leadership and managers at large to medium size providers and research organizations who are exploring big data drive your business to improve ROI and optimize patient care metrics and workflow with the right staffed and supported distributed analytics teams.