Solving for the data science conundrum

Across the industry spectrum, data is now extolled as the fount of new insight into market opportunity and operational excellence. Better collection, management and analysis of data, so the story goes, can translate into improved support that enables managers to make the right decisions faster to enhance competitive positioning, performance and/or service delivery. On the question of data volume, this script is well crafted: EMC’s Digital Universe Study has estimated that the total volume of global data was 4.4 zettabytes in 2013, and is growing at an annual rate of 40 percent to reach 44 zettabytes by 2020. This is an impressive amount by any byte measure, and in some sectors, growth is even more significant – pegged, for example, at 48 percent in healthcare. On the question of data value, research has also uncovered significant potential: a 2013 McKinsey Global Institute (MGI) study on “game changers” in the US economy has ranked Big Data as third in the top five catalysts that can spur productivity and renewal, estimating the potential cost savings derived through better use of more data in healthcare, for example, at close to $190 billion by 2020. But as with any good play, there’s the rub – in this case a shortage of data scientists who can manipulate information to generate actionable insight. According to the MGI, by 2018, the US alone could face a shortage of 140,000 to 190,000 people with the analytical skills needed to work with large sets of data, and 1.5 million managers and analysts to apply this data.

Beyond the macro scenarios, this chasm between potential benefit and the potential to execute data strategy also operates at the organizational level. To narrow the gap, a number of campaigns are now underway: while educational institutions are developing focused curricula to graduate more data specialists, organizations are experimenting with the deployment of analytics teams that together may replicate the ‘scientist’ skillset, and data analytics vendors are creating platforms that democratize data by putting its power in the hands of increasingly broad user circles. In cases like the Guelph General Hospital’s, all three approaches are being pursued, and have been combined in a new initiative that has just passed the test – a first cohort has completed a new course in business intelligence for second year students of the Health Information Management program at Conestoga College, featuring accelerated learning in healthcare BI, as well as an 18 day placement at the hospital for three students who used Information Builder’s analytics reporting software to apply new concepts to three discrete hospital challenges.

Project drivers

Outlining Guelph Hospital’s interest in participating in this initiative, CFO Gavin Webb explained that the hospital began to identify goals for business intelligence over two years ago, driven by the need to make better use of data: “In our healthcare world, we have reams of data. What we were finding is that we were reporting a lot of this information to the Ministry and to different agencies, but we weren’t taking the time needed to understand the data, to convert it to meaningful information to support decision making.” According to Webb, the tools and the staff required to effect this kind of transformation can be expensive; however, through RFP, the hospital connected with Information Builders (IB) who were able to deliver a manageable solution for creation of a data warehouse that now forms the basis for the hospital’s decision-making model. To extend this, the hospital looked to one of Guelph’s strategic directives, which is to explore “innovative collaboration or innovative partnerships with other organizations. As Webb explained it, the relationship with Conestoga College was based on this mandate as well as the need to derive insight from hospital data.

A good deal of the impetus for the new education project was provided by Doug Mitchell, director of decision support, a seven person team that forms part of the hospital’s support services division. Mitchell describes this group as “part of the information portfolio” which “sits at the intersection of IT, clinical information services, informatics, purchasing and finance – departments within the hospital that are information based. The genesis of the project lay in Mitchell’s conversations with Justin St-Maurice, a professor in Applied Health Information Sciences at Conestoga College, on a collaborative approach to BI education. “This partnership that we established was driven out of a gap in expertise and knowledge around the tools and the processes of business intelligence,” noted Mitchell, who also reached out to Information Builders for software support. This project was not Mitchell’s first experience with IB: the decision support group has made extensive use of IB reporting tools (Data Migrator, DevStudio and InfoAssist products) in accessing the hospital’s “DAISY” data warehouse for creation of apps and dashboards, ranging from a real time GGH patient tracker, to a “metrics mart” library that tracks and trends KPIs in other reports.

The new BI course was a joint initiative of Conestoga College, Information Builders and Guelph General Hospital, which also benefited from input provided by senior BI analyst in Guelph’s decision support group, David Dornan, who helped with curriculum development but also with creating access to DAISY data for students who, as a result of Dornan’s efforts, were able to work with real healthcare data sets. According to Mitchell, the key goals of the program were to offer the group itself with more in-depth experience with the BI tools, to “grow up some new [staff] resources,” and also to ignite the innovation that students uniquely bring through their “receptiveness and enthusiasm for learning this type of work.”

Implementation

Prior to their placement in the hospital, the students completed the theoretical portion of the course – classroom training that St-Maurice described as focused on both healthcare informatics and health information management, a blend that supports the education of working professionals “who understand both those worlds.” “The newer program design that we have for this program,” he explained, “is really to create a graduate that fits in the middle and has the skills from both those areas – these are the ideal students to place into this kind of environment. We have a mandate at the college to give students hand-on skills as well as theoretical knowledge. A project like this fits exactly into our raison d’etre… and the program success demonstrates how a college can contribute to the community by delivering to partners.”

While emphasis in the new course was on practical application of BI concepts, St-Maurice noted that from an educational perspective the major objective was to develop a generic skillset that would enable students to use a variety of tools to apply concepts, as opposed to teaching skills specific to one platform. However, “the fact that Information Builders came to the table was awesome” he added, “because it happens to be the tool that is used fairly widely in this region. So strategically it makes our students more marketable since their intake and uptake times are much lower. It really facilitated our transition from the course into a practicum.” According to Mitchell, another important advantage of working with IB is the AppStudio product’s ability to provide students with exposure to the “continuum of information, extending from source data through the ETL process and into metadata in the master files, to the production of reports, their collation in dashboards and publishing data to a portal – to that entire process in an integrated solution.”

Student demos

In a presentation last week, which represented the culmination of their work at the hospital, students Sydney Reeve, Corrie Fox and Hashim Garsi demoed three dashboards that used hospital data to deliver insight into hospital or departmental issues. In the first exercise, the student team used payroll data to help track absenteeism, applying the Bradford Factor to analyse information on total sick incidents, total sick hours, total employee shifts missed, the number of days lost, and days lost per full time employee, developing a “score” that can be delivered on organizational, departmental or individual employee levels. The dashboard produced three “use cases” for the organization, for directors and run time reports, various views on the data as well as potential for trend analysis, but most importantly, the students pointed to the operational issues Employee Analytics was designed to address. Attendance is key to service delivery, and uncontrolled absenteeism affects cost and causes disruption of workflow, they explained. A better Attendance Management program can encourage work attendance, raise awareness of the importance of good attendance, and ensure that the hospital is fair and consistent in its treatment of employees, while identifying those who have excessive absenteeism and providing support to employees that may be experiencing difficulties that affect attendance. Better information, they added, can guide development of a framework to encourage better attendance, and direct communication with employees to show them where they fit into a larger industry context can help to change behaviours.

A second demo explored employee compliance with the proper fitting of Respirator masks, which must be done on a regular basis to protect against exposure to air borne viruses in the hospital and the spread of infections such as tuberculosis. And a third reported on employee compliance with eLearning modules that are mandatory for new hires. In these examples, the students pointed to additional dashboard features – for example, colour coded data visualization in the eLearning case – as well as action items that may fall from the collation and presentation of information – for example, delivery of an email blast to drive individual compliance in the respirator case.

Evaluations

Response to the student presentations was overwhelmingly positive. While Webb noted the technical challenges – difficulties associated with extracting data from the Meditech system, loading it into the data warehouse and manipulating it – Mitchell remarked on how well the students worked together to make a real contribution in a very short period of time, and Dornan noted how quickly the students were able to identify the business issues, and develop a tool that could “change the [hospital] culture in terms of accountability.” What accounts for this productivity? On this point, the student perspective is helpful:

According to Hashim, “though we used the BI tool for less than a month before our placement, all of the coursework before that led to our being able to use the tool effectively in a hospital setting. There’s a lot more involved than clicking buttons and slinging reports over – only five percent of our time was clicking buttons. Most of the work is actually theoretical.”

For Corrie, “interaction with the tool was easy, and the interface made it more familiar for somebody like me, who wants that drop down menu, or that point and click – which is how we were taught in the demonstration videos. But when we got to the hospital, Dave Dornan’s solution was always to go back to the source code – and here is where there was a distinction between what you were learning and what you had to use at the end. They’re totally different skill sets that you develop by using the more simple, Office Suite-kind of interface or the developer side that the tool also has for users who are more technical.”

And on the value of working in the hospital environment, Sydney added, “this course that we took in BI is different from others because it’s based on what’s happening right now in the real world. Other courses are more theoretical – what is happening in administration and how this works – but this is more real life. It’s nice to know that you are doing something that will make a difference – that will improve things.”

For Webb, assessing the productivity of an analytics project with ROI metrics is problematic: “It’s difficult to say that building a data warehouse and BI tools that it will lead to x amount of savings – to connect the dots in an ROI case before you start. So at some level, it’s a leap of faith. But I fundamentally believe that if you can unleash that data, and put it in the hands of the people who are making decisions, armed with the right information, they will make better decisions to drive efficiency in the organization.”

Webb’s confidence is reflected in hospital plans to launch a similar student project for the summer, and in its commitment to extending course participation next fall with two, rather than one group of students. Interestingly, the students themselves were not familiar with exact meaning of the term “data scientist,” but definite about seeking work on graduation in the health information field. And with the support from the hospital team and the right information tools, have certainly met the criteria – developing new associations through use of data, applying new concepts to gain insight, followed by reporting to solve business issues.