Dr. Dayanjan (“Shanaka”) Wijesinghe is an associate professor at Virginia Commonwealth University, School of Pharmacy. His research focuses on the development of precision approaches to health with a focus on digital technologies. He teaches PhD and PharmD (Doctor of Pharmacy) students precision medicine, computer-aided drug discovery, digital health, and rapid prototyping of concepts related to healthcare using low-code approaches. Here, he writes about training future healthcare professionals in data science.
The need for data literacy training
“Data literacy” refers to the learned ability to comprehend, analyze, summarize, and draw informed conclusions from available data. When properly applied, this can lead to implementing data-driven decisions that maximize favorable outcomes.
The healthcare industry currently collects data from a wide variety of sources such as electronic health records, wearable devices, and clinical trial results. There is growing evidence that, when properly harnessed, data can be used to significantly improve patient care.
However, the healthcare practitioners best suited to gain actionable insights from this data often have little to no data literacy training during their professional healthcare training. Having good data literacy will allow healthcare professionals to a) understand and analyze data from their own patients to inform subsequent clinical decisions, and to provide proactive, personalized interventions; b) identify trends to inform best practices to reduce inefficiencies and bottlenecks, while improving outcomes and reducing costs; c) impact nationwide improvements to healthcare by properly communicating results of data-driven interventions to other institutions, thereby leading to global improvements in standards of care.
Lack of training during early healthcare education is a significant drawback to achieving such impactful outcomes. Investing in data literacy among healthcare professionals during their early training is essential to ensuring that the healthcare industry makes full use of the data at its disposal to improve quality of care, reduce costs, and drive innovative medical practices.
The challenge of training a data-literate graduate from a professional healthcare program
Despite the recognized value of data literacy for healthcare professionals, several limitations prevent the implementation of such a training program.
Primary among these is the lack of space within an already packed curriculum. In professional healthcare programs, there is very little time for training students in disciplines that are not considered directly relevant to healthcare. Another significant drawback is the outdated concept that data literacy requires knowledge of computer programming. Finally, while the teaching faculty are often highly qualified in their respective clinical and research domains, there are very few who also have the training and background in data science to provide the required training.
These challenges require a slight departure from traditional teaching modalities for current students in healthcare professional degree programs. We will outline the steps we’ve taken at Virginia Commonwealth University School of Pharmacy to overcome these challenges to create a cadre of data-literate students who are on track to graduate.
Step 1: Debunk the myth that extensive computer programming knowledge is required prior to analyzing health-relevant data
One of the most prevalent paradigms within the professional healthcare teaching faculty is that competency in a programming language is necessary prior to solving data-related challenges in healthcare. As such, most of us traditionally focused on trying to find time within the already packed curriculum to make students sufficiently competent on R or Python. Most students did not have the time to devote to properly learning the syntax of such programming approaches.
The alternative low-code solution traditionally used was Microsoft Excel. In fact, in most hospital settings, Microsoft Excel is still the standard for analyzing health data, leading to limitations in capabilities inherent to the platform itself.
We realized that a different platform was needed. The ideal platform would be a low-code solution that was highly flexible with respect to multiple applications, such as data mining and text analysis. The required platform should also be scalable: once an application is developed and tested, its deployment within a hospital system should require minimal input from other personnel. The platform should also have the ability to be coded in a programming language such as Python when available low-code functionalities are insufficient. We also needed the platform to have a sufficiently extensive, but small and modular, self-paced learning option that could fit within small gaps in the curriculum. Finally, the platform should be distributed such that a student who graduates with familiarity with the platform can continue to sharpen their skills, without having to rely on an educational license.
Following an extensive investigation of all available options, we identified KNIME Analytics Platform as the low-code solution that satisfies all of the needs outlined above for teaching data science to students in professional healthcare programs. Our results to date have been an absolute success.
Step 2: Create a minimally disruptive training infrastructure to allow students to be trained in healthcare-related data science
Once we found the appropriate platform for training our students, the second challenge was finding time within an already packed curriculum to provide the required training.
We identified three challenges that needed to be addressed at the same time:
1. How to provide an understanding of data science and digital health’s potential within the broader healthcare domain.
2. How to provide the necessary early training in health data analytics.
3. How to achieve both goals from the first day of the students’ Pharm. D. program, without impacting the existing curriculum.
We realized that, at least for Year 1 and Year 2 students, the only means of achieving these objectives were via an extracurricular approach. In conjunction with two Year 1 students, Emily Ko and Amir Behdani, we created the first student organization in the nation titled Pharmacists for Digital Health (PDH), which focuses on providing students with training in data science and digital health application development through regular meetings and a workshop.
The meetings consist of an invited speaker who is a leader in the digital health or healthcare data analytics space. The speaker provides an outline of the current landscape within their respective domains, including the challenges faced when trying to implement data-driven solutions in the clinical settings.
An end-of-semester workshop focuses on providing the skill set for tackling the challenges described using KNIME. To enable students to showcase their progress, we also created a digital badge in Digital Health, in which attaining the badge requires successfully attending the talks and participating in the end-of-semester workshop.
PDH has since become a non-profit and student organization to a global body of healthcare students. The organization is slowly evolving as a place where current practitioners can also obtain training in digital health and clinical data science.
Step 3: Address the challenge of limited instructor time by tapping into the comprehensive self-paced learning modules of KNIME.
A significant challenge to training students in an extracurricular manner is “time”. The curriculum is packed, so finding gaps for additional courses is not easy. Each student's free time for learning extracurricular material is also different. Providing instructor-led training to a large number of students becomes an impossible task.
We needed a platform that allows students to learn at their own pace, in bite-sized modules of less than 20 minutes at a time, and where the completion of their learning can be demonstrated. Again, KNIME successfully filled all these needs through their self-paced learning modules.
Our target is that students successfully complete the L1 and L2 modules. Results so far indicate that this is more than sufficient for a student to begin impactful analysis of healthcare data and the development of digital health applications. We strive to get students to complete L3 prior to graduation, such that they are then also competent in deploying solutions at an organizational level.
The short training time prior to real-world applications development, the ability to undertake the training without dedicated instructor time, and the ability to fit in the training within the gaps of their existing curriculum, has all lead to our success in using KNIME as a tool for training our PharmD students in data science and digital health application development.
Step 4: Create a dedicated elective for learning health data science
Pharmacists for Digital Health (PDH) provides a successful introduction to the use of data science in healthcare as well as basic skills on data-driven healthcare application development in an extracurricular manner during the first two years of the PharmD program. With that, PDH creates a cadre of students that are interested in pursuing this field further by their P3 year.
The third year allows for curricular training in the form of electives. We created a new elective for the third year, called “Introduction to Data Science and Rapid Prototyping”.
The PDH programs often act as a feeder for programs for smaller groups of students that take the elective to further hone their skills in clinical data science. Again, we use KNIME as the platform of choice. The students learn a variety of technical skills during the semester-long two-credit elective, using a problem-based group project approach. The final product is a group project that specifically addresses a clinical challenge. An example of one successful outcome can be found in the following video.
Step 5: Develop an advanced Pharmacy Practice Elective (APPE) as a comprehensive boot camp for the final year of training prior to graduation
By the time the students are in their fourth year, they have all of the clinical knowledge and a significant amount of practice knowledge through their clinical rotations. These students are now at a point where they can identify a clinically relevant data challenge on their own, break down the steps required to solve the problem, and when provided training in a low code data analysis platform, they are also able to create solutions that are ready for testing in a clinical setting.
To produce students with such advanced skills, we created the Advanced Pharmacy Practice Elective on Digital Health, a 5-week, full-time program.
Prior to beginning or during the first week, the students complete the self-paced KNIME learning modules L1-L3.
Second week onwards, the students identify one or more clinical challenges that can be solved by developing a data application. Following discussion with the instructor, they select two projects (one primary and one backup), where they will develop a solution using KNIME. Once the solution is developed and tested, the student is expected to upload the workflow to the Digital Healthcare space on KNIME Community Hub. The student is also expected to write a detailed blog post on the solution that they developed as well as a YouTube demonstration of their KNIME based solution in action. As of this writing, three students have completed the APPE rotation and their final KNIME projects are listed below.
Successful professionals thanks to early exposure to KNIME-driven solution development
The publicly available content developed by students shows the success of the program. We have also observed that students who are exposed to KNIME from the first semester of their PharmD, program go on to do amazing work in the field.
One such example is a startup created by PDH founders Emily Ko and Amir Behdan, called KBHealthTech LLC. This LLC uses KNIME to create automated diversion monitoring programs for hospitals. Their product saves significant time and money, while also making the organization more compliant with controlled substance diversion monitoring.
Further examples of success include students taking part in – and winning – national healthcare data challenges – such as the 2020-2021 PQA Healthcare Quality Innovation Challenge with an SMS-based tool to bridge the digital divide in healthcare – due to early exposure to KNIME-driven solution developments.
Here at the School of Pharmacy, KNIME has proven to be an excellent platform for introducing students to healthcare application development, as well as taking them through advanced solution development phases, to creating practical data-driven solutions and applications for broader implementation.